From ca5fa333ca5a2e277cc998444ab936f06098c0cd Mon Sep 17 00:00:00 2001 From: Matthijs Berends Date: Sat, 3 Oct 2026 09:06:27 +0200 Subject: [PATCH] (v3.0.1.9095) Taxonomy build: fix root causes of missing and misclassified taxa, add unattended run prompt Review of reproduction_of_microorganisms.R against the 2026 intermediates: - genus homonyms: overrides, LPSN/MycoBank first, status within LPSN/MycoBank, move misplaced protists, remove true homonyms (Necator, Capillaria, Graphium) - keep clinically relevant names that are not validly published (Tropheryma) - one set of relevant genera for all filters; 30 genera added to MO_RELEVANT_GENERA - MycoBank classification parsed from all columns; empty values never overwrite - GBIF synonyms in another genus get the lineage of their own genus - prevalence no longer copied from genus; current names inherit from synonyms - inferred records follow the relevance filters (no more unrelated lichens) - ASCII names before deduplication (12 duplicate full names in 2026) - MO codes: robust code parsing, reserved old codes, single allocator, domain column used directly (2026 rename would now fail) - name clean-up applied to taxonomy before MO codes, BacDive rows filled - unattended running: review() instead of View(), CSV export of reviews, LPSN cache and retries, stops when raw source files are missing - reproduction_of_microorganisms_CLAUDE.md: prompt for an unattended run with findings report and checklist for human review in the PR Co-Authored-By: Claude Opus 5.5 --- .gitignore | 1 + DESCRIPTION | 2 +- NEWS.md | 2 +- R/sysdata.rda | Bin 117196 -> 117360 bytes data-raw/_pre_commit_checks.R | 30 + .../reproduction_of_microorganisms.R | 1853 ++++++++++------- .../reproduction_of_microorganisms_CLAUDE.md | 284 +++ man/as.mo.Rd | 2 +- man/mo_matching_score.Rd | 2 +- 9 files changed, 1372 insertions(+), 804 deletions(-) create mode 100644 data-raw/_reproduction_scripts/reproduction_of_microorganisms_CLAUDE.md diff --git a/.gitignore b/.gitignore index dae8c0104..7e6d90d87 100755 --- a/.gitignore +++ b/.gitignore @@ -30,4 +30,5 @@ data-raw/DSMZ_bactnames.xlsx data-raw/country_analysis_url_token.R data-raw/country_analysis2.R data-raw/taxonomy.csv +data-raw/lpsn_scrape_cache.rds data-raw/WHONET/* diff --git a/DESCRIPTION b/DESCRIPTION index a3d292cdd..6e7156cce 100644 --- a/DESCRIPTION +++ b/DESCRIPTION @@ -1,5 +1,5 @@ Package: AMR -Version: 3.0.1.9094 +Version: 3.0.1.9095 Date: 2026-10-03 Title: Antimicrobial Resistance Data Analysis Description: Functions to simplify and standardise antimicrobial resistance (AMR) diff --git a/NEWS.md b/NEWS.md index f2afaac2a..7791d0f90 100644 --- a/NEWS.md +++ b/NEWS.md @@ -1,4 +1,4 @@ -# AMR 3.0.1.9094 +# AMR 3.0.1.9095 Planned as v3.1.0, mid October 2026. diff --git a/R/sysdata.rda b/R/sysdata.rda index 14c7e85d98a39258d4b1c07d1a11938b51261c76..4bb94462526bbe079525a5e3a0b13ab30cc457c0 100755 GIT binary patch delta 58586 zcmV(tKr}T}dj`BQqxAG^Ma^*~F zrtD>)b!H3X{#AEv-S6^j@0W7GT`|y^QM^mgLLQl^L>Li!(YHA#DIN$aZ$y_!+R#RX zWeUkP=CvX0O+r1qte}E1R^ZK}JHOP+44KR@C_8OD36@iv0iPe*>BJ>uf;+AWfAd-j zI5dG|A&uIMLlTsAyAn(x`6yuuD_xK6IlZ|+%ziLS5 z6!Dy(oVdhU%kD$3MY3XYV`d-KAxNycRHvu*x@=D~$71PBbbEl{_S_EYf7r7IDfaC_ zZ*+WWIqAR;g)vxKi_!;_8?5Tc*E^x;{4F5QoFQ{(e@ZWL(46J&%KWYPClPW;q5V_SxfMh6{wMez=GADZcJu;; zNFDI7)+^iA=|C`lbQ&dH+nt0BguxdfF9f9uFm(JiP(;^iK9m~8A)!kpD2R)_pEAB5 zU6|$&eb1 zFGkF8Xm@?4)`mn=h+n!J^cA#GE_aSsOqr`;&5GmC*K?hEnK06Ijn>(J*tq{@mUo~v?Somm5oDlBQ3;t-pasb1C1bEaHQ)x|ZH zw6jvqk{ZfO#>z~^e}MwS9J87Qh__a&o_&K7a7JMIGy0pWEnBmF*5WPX*rXHXU_*c-xC@G>KzY;5<|G??- z#xO_aG&>`j_8)TIsV8KF_*!rLni2ky{vzX`$+(xYGomoDe>}BCg>)%$B&62t@wUO0 z=H1&+q7K%BzAl2hfrrk^yh#bgAzmT*NbrKbg<>gQB32ZZ*qHUe`%Ls6#q#9xIP2cF z_N)A7=t>xE;R!!^0{Z3-`pv=4t9j%RH}&bSa?=O- zIV3I>c%gQ?Yp;M%k-Bx(;qR@Gp4=TYhymj{A|Vey>))KP_UcpWWf|CPZXIhYjKdt#nOe1a7$h9Bf7mmfR&@RWjxT#?x0rRYRUh^7KXEqsV10>rULD0}Be5lKp6oeAZzS z@v?rgOO&1|=p0Op@Hfw?u2B&$Y+K6qtpwx=e|}~wo(K5cm2uRD4T@q0>kyLVeFo#HT zRA=AHKR+so6#RqLF{rKxH9ZKv>64MK(OBeFP%?hEcjfCx_`HjnWFW^^bv^G{eA;{Tiyyd34|9A72Y@&Ebn>Vof2Q4kSTOFaIYFdr>x!us^)Qin8*}2x`&+9c zRx~|MgoHBjauPUzoBZCZfVBtc+p-<=G*OX}w_&|kt9iajT;>)iY3<}!b9lsFYIDPhOl0Qjq&ro_ zURi!U8O3P|Z`GpJbxqhBbo{sL$5!+2vI zFFhx3YcXKa{U^&&Cf84mN^0zJ!W6&FE{=OgqyfdX;lUN<;{FK^nEohPlXGi-ZHPo>lG=5dBScLFf$g*yB01zvZa>_Uj+iQt*pmYPCRxDeU_fA-o@&!N$` zqRGMux5J5(E{~I#102hO=7B?qig%hOBJ~h(mhVHDR@(Z-u0_hxGlx@ z)WLNGLtakNEksNM$ZcjufYhv8#;93k0_^M%_%epzgpv=uxfD!refT%2avT+@J`Kt{ z3-<3+H43*+%MqM%_dZC-f0KV{|6-pE0_oezedKbzEqn$XzI^gmcYcTYpdt#Px9b$> z?cS_{>5u96t6hyH;f}LIHWEE-b5;zZgk7G72Wr|b0QK14@I%bvMm;V>@2EzG@7X6% zAwRR!N$rT~TnvtR;TO_Xjd{*Hi9cM*(f53RI}vq1=);AWm%w2*;0OEC~o* z!zKdUN0n`Hl7l0@eZqFS=yn8P5+0SDqF5-et1AuItiwpnoy4Vy^vq+xY#JRJ`A(+Q=IiLTv-re=C4#!w5wjAe@gRL4_>yXTnY~Kfz`jlO247Z z&u@y2mLwQdz28w!G5Kn4f_2u@iEorILLHRVS~~WiBB2+PA^? zr)+jGe`}U-&yqAJ7#?mvT&_y)zEnJp8gO_6GzcY~QM=YT=4vsdTF&T>F}mh=r+_fw zsk48eRYud$*lXlUX3|2h$UZ(#c^>DCyvpR)asgSu;?q9Z9Ym_U-6&h#5SU_Qufn)K zALw}SOGU0TIw$l;>Rxk&QESs1Oj%X{sk}>xf6-xA2t(9Pzf^)~iG1w^FKfecfjD*v zr=5n8lI;biHXm$EO3{J1G=Ds`C;JF+lt0NyVDKWQj^p)tD*~>L*OOQ<@jYT@XMaPm?=W1Ll1{4oI!)v^M0LaQU7Qy?(daUd2gq}FI;{ds?qR`xT#7d(-)dM$ot3D~9 zf8+}(KM)&DIgfLLR?Z-}6-m!`wUC>UgD-mw?^fYstc8Eap(r;}r*J0Zz*gHb$a3_q z>ea7tqM8I}jV39aSum5d+ZWI^+@ZqYdWK;`Y)>_RcS0zc zs6|tx)YopUnZh}Cqk*6li4|1LL~}y&f5eFx6z?&T6DuLk@I~3K9V|JumL3PiSSGZN z^AYtMIgo3WNs&pJ_HjGoZwMS23L^&WCw~jxZK~iGm@ohq zt4s%R`j!_cGkeDck}T^Xn**cmf3O)EWj-gUy(9Jrm_3f)eRay5P%5%va_>O%!7Gt8 zAsR0`vt0=0#=_TC6(yJKrw5}5jYpgGHc^u(r>6cYQQO@>qwWmP9w zF^-Q+1NH_b9i;;EB4qPGX>}Xd}B31*T zR5yG_=L$)3ejRGEY4Z`nf2ZA*(mjex7Lso_J;{ve2W{$d+)1<$7N7x|_MLGZyXmS` z9HbGKC*_)*pCFD-K!4Fn?3!#8O9^ZuGXRE-sV?OmRs>P;ZJC`R0p4X+l?)^+dH=qZ zAie(X#FIPA@|npreo>x^Ryre)?9kI62lq5}B(R#b67mKUTNwb>f5jEI>(Q^kE)xD^ z;4lSz0?6eOfGW9S`j_CM{)1}z!C@H|7)9(6YHg)bm_|ojPUl?hmEv!gH>gdjc>pWF zK{?!vs`qkK(qI|3{e~kc)Xqr9mv%)5ZTZ&6!u|jz@ck@aCephDkj!GfA0*zMc#24R~l@iSYkoka*LE|+ryvUp+V`@oz^$F@*ak^ltK?F z%i14PrxSVQdraanYSQ3vSxyWRxyJ4hpWQZ3-k^)ae+u2G(3Bx3ZkE%=8!qK#X$UpO zE7njpSHG=aH?sZz_!;Am1#OJLX6G*81p@LcE%}%$AneLjQOx* zCrLsjn9(!o;AkOyo9eJ=+d@YhJ%sk}X3R%vZu&=$oU1!PMtdT{2rE?M(lT_4OtNTU>tMM-!P z4FbH`CfgOQv+t#r?{5euZziKqS<716thA%vp4!~De^?5);7gAq2>-ogh{OvA1V82s znv4l@j53q7yc_wNDf?phpe_MWoX!s6OE54W$Zq9?*aXAjL`d-23!n`yswCw=UVw>g zqIm1RW9tRqDfrLuYmSNyI~i)=UQISfZ4rAPoNE=%^RWyQ9#`5#5yse{G(G$QxxP8C z90I>$e=W_?=hP6pVdIn69^m?_n+)aUo78n<+&ZveT62oX%6|i0vtVP}%QgdQb_&fe zopE{0^VmwirRy%r$MZhl6Xyu!6SlFODmnO%-PK%UDV6|I-OZ-Unou6IvV}}Xu@G0` z&bt^c=FE3>Y7F3uygv%?1pZTSFj$I|K>vQke?AAhqF(a8r|QpypYh&HwkYUhsV|BH z8+Ny^hmj3$M^n7(&V7$F}H+X zXsNa!1?1}|IH)Oue5;OTjXw7M{x*2U3Jn>-?U>`Nz7~+3!jHt=2cnn1#T%6=AQz3- zf7-mqSfap(v8E_?wG0uGKMrrQ_X3C-76aV(Va;_?$V#6nJdlX`mfs^&?fPpw9dT=! z;zNu;1q*O58(#*IR_pRZEpD8tQcY?o^-imUFZMR5#~yc|6JTTf*@@o=ckJ2-;P7O4 zxByvlP7Xul3IAg=Yvh5(tcg^iEFoY$fBQnPk!UNO-uEEZtji&|@3ccRxO>>E#||E& z4jG2~=oafg88<=40h{wC&Q_%ayeMSuv_l787`|b&(E%Pdb}OGi>ztGBg;$2&RVAXc z_DhpZIQTB>$6`%Z4h!T!I*+EniV!4S2KRldN3YI`la_U#cQt$e2e&iSL3*Ffe;X93 zZXNC0D7DTP9%&$~Gwr~uTAxz{DjN&_lPQVv2ks1Qk5kORngUMploBqwYD;uJyp2vj z{RES<)ok%g^RV_U4o8n?MMX{}I8#9_`V{1?6D~0_0h9iKB{_f|=o(<1Osn$`#i&aI zi7vSh0X8d92i(4@WgH15C%+@Ne~QUWQ7#C1 zmXlM4%A*OuA{V%oUB047+V099XbE+%_mndspK9G>;FdV*X3VOiZurncKw+-|iM1>1 zTziG}Ybqr(9&`}surz_+yYjgZVoL+LLpIsoyvHZ6a;xy2+wij^qDiO~Psk)0=5DvW z9%=p2&yZxHUU#q7G#I*!PM?(vs6%X70s?HKAb}`IRIsuRq=nX}R`a@;hBfe?W~FXw=FiseqkP z2S^+8wa!3Yp1Rr3y4ur0p^s09jgJ>VieuLtLCyp%5;lF}ca!l*hdx*oWu+vNu0<(u z^7_Ztes7if9|2`~iID39hXxf~W__;*Lx%9p@(_gK;^ zAG9>%qgDg6?+rDDfA2&I-4Z9F5tEDV^`zW-B4U{r?axg%g~xyWApPk}d>)GJ0LSm# zTK&{P9_fOhAf2`aa1F0)wDej+oV*c!=~s@Na+u8(Qo;JfJ|Tn`0dz7u-cxHL1|;0& z@x$O`)0YfC<3;cCKQ+z2OzGzN({D)*D%1FJpq8&`@X5HpJYYjvT>wS}ZkX4JH%<+SYs zm+<7{vW@kNPLyZxW!m_5$Jw(YgkT6~9C>%>{r*27u$>6`+03$czj7n1F$f49UXX#@ zptTY}ka>=t*c3XI_`ZQ_S z4kA7ofCJw*|WQZF;D22l-*Rf zaNti*8(??Fs4WX5(8XlSGZgr2))b3FFD;!(!i+~<8#P_ zUNzP=XRGUk^)PNf9*We8dmc)bB=I_FP!ycgf3=RE2xcmL0^nyN+db5&s*F_a!kC6a z`C5dpai|`=lNd&AjmW4|DpTU9(yS6vMIx|{c=pZhH^l+P0!3CUj0%1$KlhzV|Mowh zf7N`4i_QJ=gyXAZo*q^mg>&g(>O(4V;PGh4k>IUFUa5AK&wj}-{HaWnwAwR^#!eo2U0w4W4^LDl((=nG zxvVE_6|UBizMoF43qvHB1h(G=PzFg$bKb^+cw&*f{I2Ukn9A_l#H8w1Uj2+$M{|s` zvn`i5Tpve|R?57KNc_?|)(~f~DN{zqe>7}9ZTNc=HN~?Jrc_3vQ5Nt!L%9veh&?Hw zkTj7$J{rHO(1p#K?DEqtc^ZRP$@X|8p9e71vyA2%JzynB+$s?N$K0&2_C5BtE05F~!wJ$L zxfE*-r`e-!FEx%)b#S_!-GwjKU=E4U%<%eS)b|xJ?aq$cYUQT5pS@3wS$$0i-*8VP|26pR{rG%|D5nV;=d1fTt##*zUC{Wii zn?S}F>;4R7?FkRz*0{!*31ns7G;R(-47TW=?iDYrPL4TxUX?MI7KyNME~R^e3DJ4? zs!Uv!vf2B^?+yJnN&dy)f3&xD)2w25okdZ@Ow9gZ82eP;a#hHNQ6OgZ*bD4)}q!f!?15#!)h_VFYf*g@h+2y{Fg z_g6pO3ryFPn1s@tZmY~g8m2W2e}@|Mri;BeB2*B| z{ytbGq2_xwE~01KF_*HMWTfZ#jHZ_<$HA&n3vEtzD3xA3(El}8$0Aux`<|%jH3_8= z9&No=XiydNaXda+^qTgn&!H4QGK;KDVtz-JtJ;6&YMa{$B%VoADz~Ute0y)*ARrpm zvouEiqa-Y)m2tG}e*mC3nG_ia9o^2Fw_NI(_98?~e8wk8LLMZ4TSS5{QtM?~Hqj7-t2-Z$8@D|?X0hBB(ww{tkN zSrEOv+Zdtju0ZLbB7IHeRZ+(9L(DsoRs3Hh23rBZd0gfve;}dFCqqxci3?x8UE9Qe z6zzj#d(WEk)g8}yd@H^7;i$^l-1Bfb;*8zo3cZATCDcNjF~#Oq;gUsFHObu`he$q& zbo?7yeQW&u^2}$Tvqq8qmB*)GvXA^rv+kTde6GrU_*QBVcI!JpJggdW-eCuded(I% z1uo{c*WWlP0!2uyO9TeVNarUV5kQ`dN{R?Ts%$4We>_XZM#@xKC8pQ<#lteRnKqj4 zF2lzfhj|(!DYtcwS@}Wq*4LselyZ^;HyP_h4?J!5ERC;pG;u6MG7;e&eg{rle_&b5GiAU9NL^+DP&}Tw&3d(!pZ; zDgVlu3eFVXWXyG+X;8`*J+o^VA;#&NpmXYqSGg6>0PKuSF-*LwX zf3}8S0o0x*KCx3b|0J=UgXJWLsi6Hhdn%Ph@36x)_$Go|Jiqa<(rC5j>^{IhqzNJF zJbkUI$gO%$wFrSOjWuJaEezeEU1+kqzc#syaQje^rcjnWf=0<*lEu^mSuJr(h>#Pe zn;?x6oBJg2!Qvg$OSkcUVSIuxNde|NwheZjlFDeE{Mgjul6n03#+->prLku$_H z;8rHplB)S&i8eJ(U=q^5F)3JkS?-7otH?#iJP%d1?$7bu6&Q zJLnGoGK#0T%ickaqb|sbCmCIb<&Pl5@gswN21^*vX8!a5^{ZLVeVEhZkW-HphqAps z5oxTJ3e?ZlUrToAr`I?8|7>x2e*wVi`!p#d_zI=J`TZvNr+}WnYZFBP|Mi z$opT?*TsYA;PYp$ca@@S z{n{$i?NM!?damxfS$2u>4phpGjwu*zX`0X^c`E;kZO02QkL zr)zQ}DUbW94gzIh^+ET?snxLLo%tQy_rIqeIvXS$g*RFy>&~clNJhP)LytDF^nr3F zo9-QRda@PL_vV;Tj;#^SQkEh$kTw{j2IjP^dn^KOPyf-4(na7M&CL=6*l1Sb@n|VP zB*^f;9#pvXxzv3re_?a6n2q4AlY3CyoNUTBs?qz%$6@%%LwM3(o0F$=^&4-~*Wjl% zy!Tu^I_RV8?3~V@1Wp%}FrATi{~iekO{$zXYc-3eKn&>kzc#njs;6!toJLL5S~(M< zkawC*eD<}r=L!XT%Gur&_>nx=Nxe_Un4LXS0+*1FN#`~j~A z7g*qj_%qbcvNr+an~-c!!XRV&r9m;f7)Ks2Gl4vj=WDyA-xII(_txCm^j4YyZV&Dw zx)S%itNTq9iTO?{W)9I3FPk8?efx= z+@l9%TqKp`kfzQH5E(LFpFQ?dN2Hq6N9;mILfLEKkrQ)Qq;Dwn5$^|iZ12nc<^0P~ zD=|qje&O=#(4T_b4@djc!ecW=O_Z?55~@d76Y6mnJxt+}NlzdkFHkpg5|fjrRc9+* z7omUYi1)|4+Olt~pVImMKcoxmniq^x;--l+f3Bmz84CVD{s0%z_3KQh@5xHYHY8UF zV*}~dzeJ{ej)EJ4kxal1~>EbXU3~XDV-46#(>qd zqi93X84dY zf8des^5m2)M{rVN>FbKHK6?AbdFGr>^@eT#)?rjk{*k;ucX3D_8x$)VX)HrJ`-I2e zgHXfwQ)%OqK$~;6F(^TH#XTRo_S(7_>W|^ytyNSp{_D)GJA?{F{JSWpH*tRHMgHc! zaj`f6!KLK!HM#K`;jI|68dA&lXSm+mf5u^@L2ylb5bU(T8Vn|3wRcrJI#{dq@4|;? zs+ToEdj+`t2)MAlU|g~>yw{;n?$B=!X>Tv*y#Z)%E!$XFcsZ^}h*TJRPTKKrDPpNt z(=9aOGPm+n1&vvbzeUmDWN{kNog+$ek`kHPgvQ$a4IygbcBIMX5zt)(`oWWYe={u; zZ(5q^@S%@5>3Fqj8A<-cYxzkbeqHLs@7$HhPWv_?VV+mD%%U(x#cSN_%OneRRoadT zuG{8rr3UK3-83T%4hC~?%+IxQ>tnzC&4PPsG71H^lUTv37cB%w1-sG(;{+%wMR#A= z@}K%)s7oucVMDb5Ticbk!id;4dyO$y-t+;Z5=N%?BGy1g7>T3h)b88Tk{| z#`Csc2_>}-JNDxP;gSISg{7xT=*K%cs!0*mC(AY~i7mDHpLq?%!*%e4f6b}cQfEeH zX}dXk7aMrt#-l2xku$UTQ%`H*!4#3k28ZG-Nj)@ljiZx`1+pcQ;d&xL=-fCp#iUq5*7k0j9jEuy;bh@Er5= z+QKb6eb#9cNCQ#0ebEA^Up8;|R?surKW`oljJWIIox+^V*nRK#pX8PFxo34cJXZ(f z7U5s*X^rp@zHd(Qe;V;1slmc>Tfaj1h>5*uSF>y_29tof&*;ji?}t$Bbg7W)S@nrZ z7N$#+Sl}KCuS6^_@!tT>N&*|MZTCq=k=#qm^67iqHeAdJABI?BFw=v05S|8tK^p3O zNA}TJZW4qOe-V(ZF~4m%t~(vdQ+6bwU)uA9%Z*0oz?U~ge`n!GdFJ2L>=`kIdQa7v zCp(qdRt*7Ms;o>JqX<=GEBA(^)Z)UB5e(duuc;?c>B|luYrENi?@nj*uOz7Igzx8m z-lrplfg}J)TDjcEJt{JvAVxE5vB-)1Q!r?{`*pMKY=x$SPtIT+32oKKgfVzU8a>Nf z4Ov>Zi39UAf4_7IjG4{pQTdEuCgX4xeQ9_QexG)-RyFA8R}us-J|UxG0C-a{m>nYFtc_i-=yu zyMILR!vEy;C$yObbeeFERlwG@nq*R$TKHim=07wIf3eh%T=lEy9vi~$j|fjBi{^fl z&N0OVoj`^M7lf)NbW>fE2JA^Z?i9p%O@hg?JbX?II}Txups{1B`yT@iko`hoXx-_9 zz%3H6scZ=pWe<0oK=y-5ofitrTwMb(+b3(7v1(t%eZe(Tw_uVllZp?K)Wb3LbzB-( zvfGfde=W+)r;=~G<8r$x#^Y%ML?X2E7o2*;5ZGfvQ}VD~%6f;=)|Fh5X4X5h8vweL zm})=N{Dg!LDz!YUja@e`obQF*i%N|;g&_=5o2%}g?OUb^C?>|@NOfTA{~S|x+3N`S zO$h!+XsIt9H|kf6MqcbnufPuy8m9_#^5VADEToR-}_1OlX#gSPz+Y+}_g&;n{dDJk3wc0baxF-n!z>spie#SGrB zn@X}yJl>|-22dd4y1Wf3Eb~2s*HSV!f3j%5PFQ5zDCwL^9ioEFqQ&}qBa4DdxGydY zmeuK)gOh<%Ch3#gC^Jne>XAMPSB~=A>By)*-LJm5-EHSmM>K||cqM*;WaZg&0SNW* zC6O4hIo6v7EmMxhaK>CAonRAEXw>EneOIO9NeT>g1d~cu=c#KG^m{N80LJSAf1y!I zFfq3+vUxdJLfz$k#~=?A2d~M6MYq%ITPqj#0RSo6L^S{nxl)$v6JQ&r73Dp zPaZ`ePdt#i-+ve4l#uxp(oaT;dMxRZ`<5TGSixx#UEUJJj`9Kzgd6Cy9{vf#%+HT4 z<#GCq(sl7#nB^ePw}m|cYbZ0W`cm(gP(L3rdqFhmrb%KGNB$4%d6Twle^A_%&~czT z4E}GY@Ue$eIBNWZ2?OLFw6K(Grqbd|_8{RII*Yt7m?$}9(Laf20089ie4Bl{bM6!AbWc=iLWQ1aFcF}<9y9O0mY7&b z6>ZSJ4qJjehY`Jsg_4r`IRpWH7YF^oYfcqNxC7 zzvsL^r!65wyrp(-X7b+dpUd{Rlz^P@sqg^b!*{&o8#`=0{(Gn5|92(03nd+5K zhgRrMZyIc$sx9Li#z5bxWAUNn_2XlzR~?r+{PQPhnm_KLO3ov?f4z3UBSYy)2r32w zA6Q`0L*{J?Fj`sRTz-F9BRIDQXBDWehs6AKqll;9Ie2dri1?Grk0vvM9t@wlg=n$W z(syQ@PnTNYzUR5x56bw!D1qr?td4tg=Zel-5ANgmg?ZHg&7`ec`a4@6t!2QOEo)eL z;cY?5m~I4y1GSvgfB(&#T8Jg;%G4YynG{2c0oCKGrN(=DI8J+R5x+t*t0I%H(1Wnd z3~d9!of0H-_pbj_6RoN4?->ZrOq^@J4nh zD~Ps{ak3R5@PRdsm=<`%@|HYfxi>~jgV4I#{wxo0wMDF`1ytn$Y?`X4DP)#jP95J#NNj^F|Qnc{x;2nGJfCpnP`~E)UhEUe7vrABIhAL3ouz z8t`oG8@C}`oq?;?WVUCM8M38i*vTDrbh;wzQWz(Tp+{8>5}Ezur~$~J4q>Ep?0Wov z_3spTe;MbD2Pqm+1oTn^-gzs8e!AxP`e4H4iJdd73ie&|j4=DWgn?A>4_xjmgfH8a zr+V6(z!w^{fD2FCnVbU#|CQLOTZ{V0{962B>G;3g?{lSTo- z*ynSelZ(zvSTS@eUDfW(5m9f5*G~}So?}AHOv#$X|6&6F7bIlbLU(wNjCts>tk9BQ zMNw%{$y}RVK%9$h8g(DWg6Czl9%SI?XY=HSOukT8c;3LYHV{^51_P(Is}beEo47o99y%E7?ZyomuuUo%LLK{LZ{w!4Ly>DSNh$gdzwdekesC zM*!Q={&UXnUqY74fP2l7gKvyQq+k(JM|zC-S^r8t4Ea9p9MjB=f(D)nWOo;zF!9Nh zwu%Qm-Wyk`DCM^f1pu@~v!WkRJ>04sb4CU-4fqH9-$6HOBGpv#gA?Uaf7$$Egn=ml z-s8erg>Ab67d}W5f*T(@D!v7KYRM0uYXuz590^UKbb=WZ7@{Rm%RZ~SoDN^r%($x$ zC;793o51J{)r@kb`lczx5QAiZqD}Ne94+kLY)lOBm+cf?KQdYYmdwBFVvJ)yhae`7x2R6RD$tI!&TlClEGf) zFpWW-km_Xb6j!fA3UcNF?W~o85_vM(=n8M*Sr@kyM~Z7ZY`jFlf9)4ePK)Zh!csNJ zluxGr`835P9X>_K`%;_Ko?D+F;hu*SV|W^mEtd7L*8(Bc-x=*nycX}E!#X1KAMa2L zQ`9n^f-v*2#NjORV-(E`)YcI$Npu=Y7&sZzn#MY9#+BT#CakaTw5a&&U}TBI21MdH zGZG3-7K@HAlundKf5`5pwE^*}7sVK+=rSY7QkbIYQk8Y2XYdXEV(cha9af%&yBLSs zFSAEu2q2SOH8ba=>w&($umI~Ons6IkCxTn@q3wx9pc<3#4%K39U-l?A)$Bgi#I>kM z8aXUn7k59G@Idly@dKSd$e;StJR#DaYKPh|H(NsY!;~Tyf9-C3b9XvqzS>7sd8P3N(1^s8u$f9@N~uK!>wyZ=PPN z#?TL{y8q?P8Dxmh6fB$16h&Z@Pw}itx0fW7hM6y$DsZ*%Tk`>_snk52!_Co;% z#zsFr5p#CKWvhQ2l_JUpvl0V_l2Y()qGF8JS9no+xug5?Goj1d<*N9Jwj`>Fb-{kZ zVGPhKe{bIsr5W#5$sesn zJ{KiuJe1%~OL$nPFyrdfqAcrv3h?!zurNl4fAII79T}sf|0wPOuO&oU-yRLe=xX98 zmK;I{o2C>?zbY6}z1Nw7$z@*d@hmj&fK#j&uH0nKKTL||Cazx73ShiER{Qd}#p3b| zwfj>jh!Pd31v{_W6-T~A>8mpF+J22D2*Lhh01k<;OA%RvaKO9mB`#L-+U7xv&L>--GE0xX38w}L zEKEa?8&o*MyFT)OPmNhI@4SGuYaW(2u_sV{Kv~9jD}jj0ng?f{ zl=pYM-e@E&f7{a0XfA&Rf1P6R+D!ixH)&nCHlqrF7ns_X9X#^aBI5mP z{1OOjGV7#jx0AWr@Lnb{veiaX@~NE#k40{Q=0o1^%J-uTy)qDF9SKQt8K_}*Blmgf zB@I<)`{y*d5-Go~7ww$?q>Y#q4X6)l#9d{}60V&J(22Cgm@-16U`4ymGz1~fIsU&0nvN8 zc97Qs=M{cQh7qimHu3B4DLxs1Ly|EQA`^(s0K289+pXb6ZcgpD7~^^8h?+5kJo`Q` zJ?vE?6rE!WB<2R4LWTEzfc9hIe|U@YfHhLwe*Py)8c>3NWt@YZXJhWfZ$Iz>=J1U?1+eF@!pe`axI57ql) zyKP6otuiyeKM9LMCQqkF+_tY-&~2!h9v!t}66y?AYRpv~A~2gB-_5eg<1P~N5|UrR zJJ51$k}YR@2GPeQ+{y5J{tDMxmvTT9#%@cV_lU(F6h-mE6VsZYQFv04biEc1glZ5cBv`dwt_mRfNB})EUK}Nt z7R_P%+2pMvir|E1{SAamT0|yeT?vJ+Plqx-@4QiBmQXkYkZ!y%FZ-$-*6U#Ux=yY-)BWnJXr#zfj*mX#C* z1u8wg4mGu1O~R~ie<$M_Eea8Ofqw-O!eu7vp~lwj)MW^TPPwo)rwc<6S4+RX3V>w9L*umxe|-<>2`4UO$FYfgN3cOZ z%AFq4-B_fYEqi4ATHzF3-aa4GP2?->93{{gw_KHpsb+mab$)82@heubTH(oS~HgM}2y=(AbIDmWCgbK;3jjipr$%WR4( zgrmCeMt34rFS|6{6e_B;>9e;4J=Bz|n3%`1kxqa4qxmnLKx zO>QwuL2E@7C)8i<2W`R7T4ASqP$8YuWdJ)WFIa_7+=L(YmI#sxf)(x|;6m1Z1EEr6 za#PY>D?~fP9xPF=tXaoOFQ~g>?1}K2QDhKB{6P?Pk4jek7uiC#YcR@*^fkhIIZ_p1 ze!*Cvh z&iJL6Vxbn~NIrlSqYpsx@RVYwkI$*lR59^&L1Il<18XBaXe`3fc94u0by=+3w_{6A z9jJIeqY!tuN}1;b%1oDz=3pT?)*w}Phf&m@e+d9DK+wNd$0Ebfe77+J5IvTEZKC+) zZ1i3IVYwAC)U^G>;^KE0Y)`zPh-K{>e{C(BQvXd$`*vSbsyt?>-S&3Wxx!hgLXvt2 z@`EyFR>2>|gFQahohXb!y7XUvPGg2wRm-TRLou_lrfLKpKiYH-c6xNV6W%xgWIA&3 zm*#pp27hPFU8xez0YND1Y_bed>hUm)@nQ%t_#0I`=_}dH!kGp%P_qixvz{2=eb}r6 zkR;@T2cN&_0>IV=l=#zXX25Z<$|h(iH;@lUkG_-69^uVw=+l*k%K|ow#$?9ZxC{KU z)+J5Ynpq0VpJeuCuSZvtPEk(tPdchcz;Iju`GE*6^7w#v*iLl#Qu<2*A#u^ zqH!Uic~}PxvKYiFltj$Nky+Zug9i9mIJiWXkghJP3>52?e`E4*Wt7OB-ERoPm$8pq zT=|;CC+OWF<*qz@?AdGj(ypPxpnO8%xaeavWpb-x$!Z*Yz>X^kx)uLU-6y+;Y#Z)r z-G4{D=^i+LBt_5Ijw2&{Ua8zc-@HUOE02Eo#Yd&>OC<5oYPP799hJ*zi=K;}$(kb2 zB6kyn)80>nym?rcK_TeMXg{cOoJSzFLD%F_BGH9ZYD$aW0;;$jqjN^d#z`?ATay-C zpF!7R?qeoe8rdq<_oMM>h$wqPVtSf2et+#?BGM==CKeqgT!RFTWGy2iGTh}~?R5I8 zRdY&Q_Gsdta^h|n{Q}PCmD@*sDMB1g)&=;6H%#`7$0y`@o|=NPunOvr8ua$74V`;n z7yqv|FCU=(ROyiC?mWKZ4%W783@Eq|kds+9!2j3=3E+#NHgSpjKSa6L-hcCooqx9k zQTm2_KS5w~l)O41d#>#rxKU+v`+%mwBgWP<)@1V_y1t%y0Jc4E&w{K{t501*|ULYQ1A%Vt=*)dt>;%AJpY>6b9(iz)AWG?s$v19wK!O0D*@rDWmvRj`bTd zyxjA5NiEC4pz)yR$@^s85$Y`@iVe>JGSFI@VIKwY>A_ga5*1Ps_kg|)*1RTG9Wl;H z*3*ZLH`Tw?ykfId)-8H9lAomws1V;`)0co;%MRz69N8(1k8X^|yr*H}2hf>K`GeMp~OW&L3&n*R@|{Kn@voXvY`x6ED00e)Wp(^WnMfa``lV1~no zuO9^!8GC~W34QJ#+6d7VRM2#4=o4wv#KodWYyQH=bi-5NMJ8%{zv6E&AiEKWQL!_nC5KL~C< zxApi)SzRXXj8dN7#Id+0oX$Dw-cC?cb^=;PhcXDJaZZ-rbbFfh?9#`)7z*W^ShYb< z-7qEPoL{ky8!F;lA)~1+vdNS58E>-dl>+cDiEM^Hl|> ztYPXbF;z}DC5=Y%6*xg9w~lw^G8>vp*d}uj*7p>@gDGe6B-8?dKg{rph;b~*}23YMAT^!{0Tj+9$`>3DG9uQ3@fg*2HQ3)U>&m6haEC ztmp~L(OEYz&K5!cvcVFRrEw3vjenj60$t#as4X)N*F*UX;=K$5SR*^%Qz?=_RCMbJ zvP-mESS&Mrt;G!(9>^1oEGJfx)EN)UJcy>*6m=Jg@JcWe!-{c<2H*j2L}= zBl$H@@8Npzz009U>x#k^Tp3|SB{N9@Y0gooqfV=Q9cm+L^0-Tq&(04 znZcg{&8aZ4fULf!^S3`}RnNhulV!uj<^tu5CD+AxHFW7=G<*_@I!nfK>!H){;H9xU z(esJBKBUVwNyAA2`qOZELVpDo&9rhGmM9Kb5gyLkaAXK_+d(Vmaq%Z~Z5tE30p+PM zo7Of{kMmj(8s8MzP~VyRt@4_=V@>ORi!!v)z~l>RMPzEMSUqDK$(+9Wm`2~W_EZVxEp)1iucqfVgyqRbd{n^ z){uOW+ze6I;x#;@Y7UeS*;L1(rsf1?URzAyBxEY%r@c_u%2I?qL^YDz;jtA;GaUrA zRq#G$QGy6PfzXVs@_)Gz%M=2oy6|CC0}(;gWHDB5=rn`C@fe$+DvcgJwye8 z#|*yN^)>m}S*v)5zv7iw-+wLIPUxzi#tH`LHMS1q#*$`01~KSQCUqm!d(H6u^27^u zX_&(LjHPEu5r4xAX4D%(PveFDAyt6E(-Td|*Hc9q;k$>SWoT)`F6PE1OuLN;&^qW< zk=@z$;{f%zvc0_3M-MXW72MMFOwj{IibVl}os=yy$)rnT;@FVhl%V$$qwG7o&#a2C zi;jyTbdMHYITnK0NSvd1{vu3a z36{i_!CLEoh%zI}($e<%-_}GwmkuRF-sQuF;A^v8qJc0{s%R2qVJ-b2o-yE^Q-UlZ zSlJMwhl2Y}csKY%MmG1CupsDmWaougUDdW7OwxPCmHz#L>LBq1lB69G{nb@h zl>CQ{j(`2Vjbw~o*GMtz&{{_%pvL;VjNRZhe#k{B!eJYUz`UX{)2@f-p@F>@3f1JVPSARWKc7C?yijb202g{&|_zk0Z#6G_*#V{iuyrN+pa zAj39~@*kKXj3z-7aT}(`gtbBfC+*7>MfL(&hau_}hq+Cb@3FNu-(wQ<{jJgufAuI? zEPrY>AnEihbL*?8D`^Iu!P@*EPVE-(_oqEwKDSK(*HJ&e*dq-gnKxs(x9i1ktjn(^ zkm%F1r@XsiKham|vj;qZ2&7j++8xlDZtUYl#DlE6{8p>rwRKfXIgkYv2Nc_7^`OVm z7wSJQ%onvpfKW2%?yA-LR|ig(X|X-~k$;*Q_yC#FgGFUTeg-DWzzRiEJ#{EUoYM}N z-iCNV0Mw3p=N57b?o%#`%pXiYwu4wVo#f{16yxOYrP+}jaats>$MiI>;TM(Xz9P9+ z;X_1FH8xDSNkAV&bcu!aR7pbV(>zBm?K#2+O3w?vul&|ywEAPOhLynwNB9%PpMMFU zO+iA7p4#1cq}B_Azq@B-d~({T`5dg3ZbCm01LqsY9^bnUbAL1SY%OMGH}U!g+4@^v z{i`t1CC$o3gZD5MSDu&cdCJ*DB^jA}#_cZ@dep=yY?6)w|3YZOBtPn$h%nU5a40RAiLadqr1{i5lFhpyRvkxtlY#18rm3zC4)8y^U(f zbMj~qJ2zDy`GZ*?ItyP@R+J(|&$y3k=>JyEM>#0T`ucTQ63b}k*|a8Vi+|KPU%@9Z zNT#7UdoIadCM{XWk^t4+=s)!;uw zXF^SwEGMLiWXbnLFx@A)Y5j)FWi>iS^DQ<0$$V1q0p;MXr-G+iuJ*Y!V{spe|I+^F z5B`l0IhmN-A*lr}@qJNHDt{Y(ekl^-@JpkeF*uJ0(1GrXr(Xkcb$krfXoo57`cCcKgo2vLDkehQrhhw*M(k8OEew z=dc8p^YLI!7>Oa+@^F=7AumH_;3)>SntX-JO&7##rO3PwdWGHSDyp4aljQhMDSW9Mf&O?si z%10#-&4O(bi+@--`wCG5T{uq#)4WKx$DV9o2&z>jt7#r{{Y&8^O%W5Y-i6zCNkYcbCB++-eJTXM74GX7VcfWko zTCl(0wtoZ|C;9(qR~BZuDc%9a?jYqh@Eb$K{V|Boa?%L{#IAGpS+vTs%hY@e86?2X zkf5j8DUejL$*+clH5j0*JQ+P91k&;!meRE564uayq0$u5rm?$Wsr&aO`T|kwxqmxo z9O_l~BCw_dq!#&?EtVUK+uyZ8Z6pQ*T&Z>v9)AgHlEyXCJwVWHD(LhJ`|57bi5IdM zGT!C)$y4-EA*{oFvLU;zOO7N^qS{Qm#1+2gf8M$5t18PJf8=?*KA9R?LC8gVnHD7V zK*b~%_*&U}Due(JMn+DSFlc6blix!9s7mppqGTr=F~NqDeD|N1((%@}HmjP1$Db6Q zz<)%(vao*tNNt84_aajz1V>hYcWUmFFri?tk=kO^o@lVpR4n*v2xsS~esz_8PlzOm zx%^2k_4c(^&)`08eqUegOcaed1lY;CY%tJ0TW0%oF!yU|cWmS&IxAF} zUO)v(dPKzK|2vJKHYt2OcUhjs?FPiK6n}#mDFThrUEIH3!fv<~np=XfW3SKFoL%6A zH34vg(()9R3lDQ)0twbM@;^H?n4j_L;v+UqHds>3+P%l)^)Vu4hQ^c&2q#rbRdB0v zkX6$briN9TJsc5d!d;%o_$f?0=BLuNBxz=m!b7pKLW9%ADvoh_Is(%`Y z0^P^auJXYan$mI0b>+f}tUofTFMAE*2{+00!AUovgs4uu72~y0jnnEZ{u-fpN^Tb~ z4ADJ0cQ9DeZg(b-B<$E;|IRwlX+IK=n*pZY*i8}pw_(YCv)yQ@@y21$jMFbyVMjzH zWY)|}qJ6U@W>HUG+MLB9z6Lg=KYxUUMgYC9J{uCF*MOdsWffSa9`Z;qQ>laLY3yiw zvm=zIPqyDwN}Nq;i0X%eR)BH9*%ut7K%Tv~1_)G>MXNuJ!UhlTcVDJ*3PIo2mIveo z%=M?zZl%hDxuVjsI2z}sWgefU&f2mSJWziSUmn|bJHQ8cgq~Wy9`o>oB!8~WTcH0t zEtS@n@Z5spuwBXqAE5nL!I*P(y@A6LMfRLnza*8+WPwh5cuhhaE|Uas3yuL9b$16R z#Fk1fdR;JB@*Pk*5^Vi*gsW;qnHv2uL-E!aQQQ67 zHSA$k2|N@vCco4^LphoFpl!(#iBvImg5Lhj4xK%OBu4|7*bD$ zbs8igE|x(r*2uxM0?eg#`y#Gm3x-M~KG#t0=-0*d0FZW!^9g8;{eRH(%7PjoW$sh4 zHd2Cj_D+H%Av|qr48@s>z z1Wv!}Cg9m@g;qf}S*{ym$**;XufG`U=^hjTdo=2gWHCKrsL@@~b*lg;q;F-gtM1)B z9Nxn-6+gvV`U*mlE`JePY=*k4ExMViP`|DGv4v+7U;>ieec^bxi{i!b`s+IT=fO#+ znAu-A@2U`PWOELXh8ss(6WS47gwit;_1iw&;l(8c$zoKH|Ax>3Xipm2+O)lgvNc$% zA5i8~ITj-ql};QcXI=4~pF#cETgCqOn+$wrwKRaCQ&B{eILOur z9B8wd0Hmf9MStp-d#4iLI+>OyPo4p?Y1jDcZl^uD=*ooX`n?guhPtYhc>=gz`zv}4 z4ApH(msU%xi^o1BMVx4>#Tf2 z0kI=y*$eKrO4l{>9s;CRc)>fbJl9>r6m5VWQ;dQ@fq#?`$hEGrSoOV@gIQO$xaSJt zZ8U#-HGFHLRkV-GjiQB?ueEqj&OgA+x zbKkOzf09p(@nBlSP?hjuOOGus`PlpJBg~A@(k!nNkLi(jBV}AS!8FEYwkP@uV>>tr zjhNItB7Xt%4S1_$yapX>LQT*(h5y9^fBJj?~d}3R9N_14h=74wlnzz7=*zl z^|oU}otV?NQm#Z}Dy3j+;%g^@#-g*0*=(%#o-PaUW79PXxuvpqN*YaV?Z|1>E^*IT zUtlS!#NwlYuq{slf^l;EJ`QyDeLb#=slevE`+s9*X@og}q)N)FP~RfjlkQox9a+td zO11%?wXF23ame>ZwlgD|-@3D;gvTt5zb3^wfk@6bAn%SfN(IZQ6`=7L~ zy3A=rcGJHU-#_YduZiuDZ9rFz>F*&#R)2iwa#@7a6R-@-$X#-;=QsDi{%^N6i{3J* z9#ar!H?Mw4uyCwzHo$4!hz6a8Y+*rJx^8d`*6POIitzxFmBD&*E>@f4&)KcQ@6DnU zcvMut2r{>*+!1_1VZ&=EUy6N(rxC9w#KUkdY|-zL)Yv^^ab4?o^1)lZ?Jjd1cz@iS zPYo{Jl>LO81yhv>>q*$1ps?N+tN>D`HDT%UhFeUrNdjJnd+Asc+CrCMjpR0^MSbhwG%;%db_T#|Xf0x7l@ z4l_7hX?!T6tcyg)#xO-9noHetOyl?m578|oyWh0$4n zpbbUt?;?rZ#KtmoZ}V=^THD~r;F`>(`DK`3#&o=>-+2hr?wukN(2X>ud<`jsC_*uP zZGOBBN&GmjRnR)R6jL<}CLDt5FAOgyb?{$xPkG3_!F@X*o!BP*M7%&ES$|mIU)S|l z{FP1K66Uu^>(3G!1*D0-xPgV_t9jNx~{!-NPmHA^V(1o@JPazFeUHCQ=hQU%N>sZy3%c>f>Ngx;aq!v zom$zf{penc0Z$o}NH$#radBLn4%XFf!7&^af z+WOG9dk0ZF>sPUV3^o~0r#lkO#IZ4DqIS@Hf2#i=%pDX`5`M~-SAVn;wvBxfD}&hB zO$|K`5pc7u#-0SvGu3^dY-*)ONBf8&9hQ_RGAG?TkSK|lc%H}Srjcs_R{QpS1ZlS2 z&4yxfw7(7zP#%+lV-2SFKANXKDAijYJy;)ATW1V#TXb&VEyXd##x|t5t%``8#~rTB zoF%PGyyMJfHc;}m3V+e=K95j1mJdvQ+hb{*mvs!`o5#u34)CN=q=gadnXYHjmy6PB<;VKRb13dHpws z!gX+O8*}dqHKO@eK+gMybn+nBJKtE-z%a9Jn21(D5P^ML&wq5#*9|I<9j|~*;>^4H z>A8H>ec>mS7j#B9-a6t!$d{s9&irMqp)ytkZK`y{9^L4U z)I2Osf(KjZ$Bp?dHt120wPpV39~zTCvSe8i=Br*Ut;HI%c@f= z@2i0iu_bTp3V&#~Ap*fU>oF{Kgi8X1Ow3#})rmhKoc&4!YVavzMTev+cYq=7q;dJn z^Nz*YWLUfP8WV=)zd5G4S_Mbg4Ux|0HYU{8mbQl$Yv1IlvN%yyE8W_ zoEdJl!*xi2e9iFC)_3se($@%$;b&evu_3(L=d?LsmwzC86YJ6W8u8S@fu|ykUkn#H z?J|kaIkGb&7XKCGrv*tE9Imk1!oU9Hc6GtDsXHA{R`{+RU5cvQf`8d;UepsNfU z7n5s6V)$OZ5EK~r0Jm|JA?=%!!bHF?l*%vR1||sq(0@wj0m|+Po4&`nmRZC5i*)J7 zhai8+Xn#AO{ajOEND`TLb^k`bZsx0aBVi5zI%zNV0Z1<)5`MBI z$6->yS$~E|C-CBbi{=4o@2+|i8qI!}Pb245*nR^(cQO*7d4Azro=o#TTRXGVFc`chGJUk9Llx?fQtw8vf0LNlcJ(Ug2 zc~5)Jps$K+0!iVu@zZ@~4yVRaFr9}U(zNe+D|?ICk<=1;K2g9jvwz=rw=ehEb4e81 zBY$AZ)p#jAoaj^2f`d4(^G7qnTV-W}^455=Hz4_^o6LSZE8jOEbI2&QKt=W6;?<0+<)y) z{bAvmd$r!jJ*KyV0eW5{AqNBazro)EQ8nUr{OttLR6ZPG-fwJy)SJ<4O<5SHA^VEn zj8V%}{E~rY_@;q)G%%visibIj=b|b`rrjmyK#o|Ra9mN^bNsN`WmVoSda+`3Ojvpk zz5Q;2cYz79q@A@Hk(;gP^3+(~)c z*>{+yc4F7_WfV9$jw9LO1OH3@qjp4w>h=dw9LY{C&Jnosey{t4CV9sQ(|_ZZFU^{K zpP{}@v?r~Fa~SSXDt)?WOMmimRDDIUvdG!$G`5_O+vdd}vv2)VYFH~>eHrOi>CirF z!Cv>bT5K6D00dR@mCn)2vB>9@QzsOp!5eZ(?ljt|(MroVHE|C_-3od=yfn8!L3Q#x2fVWAt!i$LJY-ZnBl^B&9CtR>haq@R znm@t0X1IAmtE0_d&I;Ve(d;ArUw(_lye^O%;#z{d4VUY3z2hmJ&wt#nS&cZ4il1w= za@e)8;Px1LQ$sHBV}AfH_iziZvPFF?-&3T+P_iX7&7R5p#UZeI-&Yyb5<8^}!R5*s z>^$OCXF&15qmS%i4EqUeZFdT;ostC?RYP|vABk;dx4ehsA?j?m%&9mczImzqeb80w zNSWZp*gJB=mL%qbfX}0A;7xp%kNS-Kh4R97ah0MeoQ8!@6@POC!D5eD^d$XTQnC9I zK}jl2fE8tC_QNA*W2L-6$pKTcEhpPe}j?n=$5j&rPmZk+0bM=yZr=8&_=ZkN>w;o?~5ZY%n?r+ ze^Oie%_KzIE1a

KPgA5_Dxi{PoQ-Zig#8`B1fC)Z-|KqQ|eDiOFSdmmk=Vn&<%z zAhaE@o+=NTRfH=QmE#tsAEEKHa{`TI3Tqg3aJ*+A@*1y>1u| ze}9CX^A(iT`nEgVS}J-Ft_Np!4sv{icuycaP*46!Q7@#FfzuM*!PxYuD^q{rh=pz` zlbq4LAs56xpHvp6CkpFnlbD@E7hEQaPi1crBhsglXQ0co6wSV!m zC|<2mFISoC*xb@m7yyP{Y?85V_HQxvtjgb1)AUq~o(7xzi@CNy7tVO8nuEV=4XL=4 zoH>8GXEoYFKHQK&%br^XybPlR0xJJ0l~ho2T}yG|cNl5;w@lb%h{Lb|PAG&0HEloK6oU@VLv>5S} zhLCyvnKq3J+$^<^i9{G14o0tBO9@)L!yk$=FXA)+xY$xLr;s8r{G#G z{_kzdA%{7$yn}Eh%V)YM)GjUOgX!6#)-^UF&lOz zu=)e9A0VzL>`WiK+Ecn^!O4V6S{hwmfE4drCywJKg?Ti&Trhpm{!PKo&u3wSoqQ4{C1;u2a_!t17H&ux!0F*cZl&- zgn1{Tr5QA{tp>3Z*nd0)tTM)vh9Y1x(Vx;G9{Eb8|10&`C@7l*ObW)f z6`F_K(L^54`8a-6BvvJk+Of!M;MuX0)l;|p9Z-z@o=H_s`3*kV(vdatGqE3<%^_LC z0We7ScN#Qh${x`%4bYOm(*?UiX4zotIy}X86j?B|r4U3HO@C&zZ^b&j5XBs>W6l&k z8z~jux#rtl<+&7wsT7oyV1vN~wLhUomp1OZIkfL-*7^ddMM|m=7PCv-?EskXS$2Gi zqjfLY|1eULa?T|p)V9+?MeSo&EjuWZ2;yCkAFP(8flws*Gm#^osu$Yk4C1g|9pANE8A4H4J-}-QZE}D)nZ=q$b8~z zmDfg~A8r|T0q~^$8cHN!-&}6mHE8xP=9j#EZtxyTVDgjb8{}T6vmb zf+B>-8D-v)=k=IM4DJdV?%-ga7MtzGw{#pYVjU5vpy*+{q!&fAz8u)-16vqSOZ`bO z6X4fAfVm%`m-X1Bklr=zK|dwl2G4E>(s`StYkx$bmH3<2;Vp)SEqEwb&hbm$pdNEW zT|k7(V1Tv9vfK16uA#R(_#iRDt^+7 zSobRp+?*o$sG6$-Q;CH^u*$+#PH;Fw?}q{^W_i@FeT{i__FBeMDm`NPc91WvQ-^Av zZ+~28W8;=7v&NRr{PzQlw-`~C7`?xP^W!y8_T0e6ulLmbqsbymE4_+ZgWLF{71lXK z8ndP+{l#O4!C*Hz!63n@i2~rv)fHR?Yja-lV>$S?(pJDV+PCg|bC zSGy?*=d?>3p-r2{8hV2giJaqA2yny<{5&Xw%h32MTB#0i6lBD72%YVIUf6_4I(n7_ z&BtSmxi5T>{6K633fm(#BPPyVXMaa94TAt(qHSLqa1??@=zNwRKEJPJb*_CvHt3h+ zb~Z+Dm2|cDee*I`zUgEF-qB1hDF=Ng8Z^5&j9FxZ@VeA9**&VQ=;hozNoh^Tt|#$V zuL;qitdngJtTFT3j+Pn-dh>+My>;+LH3HbaSw=$6M>(G{1^4cGN`})LLx1aM{?1(C zQ_`%ojhbZ65Jf@HZ5Y$M;~?Lql-3e{;|P~`n+-xsQ4sY7L`9^ZqabNp6JV=V_n_j# zDirh0&dw3xqxvt~2+9BRU$UZE%qxsN8wkz_1^C-&&DJu>IAQ4oGIRaAS8P8HcYKcs81PC7AS7PvJ z&{hdjA2n&T;GD%BzqmXRIHEc>(C1e0Rb~T5th1rVJ>;~>f9v^2`_|qx^vOkwg31(x zJ0-tT;cOl^qqFS6nOH`)(c&hUZ_yPb-cy??KI%K9L4PaB(%@ymCx6tFeZELuJ@`kK zg%@X}=ArT1ZlsxBkBo)Si;^R_>sb?ONV%ci=vac>Tvmb_gTU2BU?1>UTVzS}ab!JU zkM*9`(RNH$V>uOpOnQR>sct_N9FDBjkYP9g7djFcX(G>Z9)}_N_?sqCO9RCv~ zIR+Re!bZi~OV)CHR1_&RnOAg0UGIfcf!zyC?Y$_X1nHh(y7YE#Z}QLG`!kUMJm zfJ21bIiS^5ju-4wP;9^$d30Z@oIVR;<1e*0jigxid{Lv^SE#VBur|%W)M{a^dqeEd zy83^(PLo-UGtry*5Y_UPGtRjml3%EzG`~Cv2P2_s(t_>*&|pvAZ#pb7Ty>s99Yz59 z@02XcaW5*B9emRt5Lc+BwaDPBR=WhcYLLGCGul+NL zU?JBJA)`KfMZsW<$rG>?96$oL*tIg3g^_W?}5)w21y2m>pB9U^S8 zgDN|CAlO~nyVur94_?zQc4l7&y2jxSrvssL9%|uj^b-tj-dNi|WPfbsvgsxPC3{V&*0o_R`o5sWyKMO9 z;3N3lSh%mQ)r1kaZqCPFqI2+#Qb6WEX3ZO>(bjuoD8D>@7>`kBM)2jlaR^_^dFrgA zktjQ-#A-f6mAm}*J<()Z@#K-aFs`S6oPbhIo{$!`w-|?Utj+#TR(naRS05LR?q|}( zSZD*Uf$K1aN+cOXyUeRIt3qNWUHiPkl>&wcT|MpV%K98*zOm z$597aL`8s#V;uIi`{uU7$$%nEuA)>=F?hg=gAT&G7(BHw7VQBEQkv)8i!b?>{(q8^ z0cg4L*l9xw$EPRatzo>=(x=yU<%d@}ZZA<*@B7iR-3hNeeS<&Ldw=`j(5s@|*>dvy zA6;luo|8p^(;7HNNt$6;8+=!*_H-UdNMn+HKG-2@BjhH|E=Q$s$O%`@^q8B`X_yCt zntaF^377wf=%xXz5fvGYB;4?4zke!&76D`cF2r&n4s1P8=zq*U7i$D4 z<8Ij}$HN0NH`ilIxflYsJbHl}aHqz-X<;}3$;CIq4dy>Feq1e~n+b@i`R6+(?5}h# z479O(;XYR3isnh=_XcK}A-sgQ2x5Zcj!p(5b1hZ$lr~U8XP-(@Kxr325x=&FCwFmkejUdDpI!rcUUci*#4J7W9gEQ}BfxyVsFV<%b`eSgpQ$hbd8_-;UE|I^r?XFt$(lB`#D0}{Lh*E))wSi}gOTry-Y z=|ta39GL@P9cCC1Cl*n{%-#Xy#!7906Z-TxEFu3s`He30w32p>DP!XTQm6z71zu2d zx`W|`JKV=`l0EA}(5N7a;WH9{k}vC<{m5^+*OeNB(&K&yuYVEFgbh^HFeET-U+8hKRUlwAwCNy(7c%28OhmY)d{n4oFV{yZd2Jt7& zoE_yY^_9Xh$$!@59we>k47_VatByk-S!L!7lUW)d_wU8VgMJwSg=nf1e+nelJSDU+ zl|EpcttRdJ3F7ny`dS(!RGoqp`)dgj&cQU= z3Ud09&#}W&y$(#Q_3bD)*>XzflqMcZ1(PV=U#m3|-hYv}6m8B>oW40v^KAV-gUgXG z?%!p%R!`qO)`a2>9lXg)aJc|(q+}nELyUbf*??ZGV0)$0e`R!xUui&f>p#z>l-#jt z+4^vcYYmtQ8qLW(bwgnJx_81u5T}}_eZwvfmJWXX{=4ciGoj7;GAJS`eZSwKmTnX( z-40O4w14=}UZJp*x7XP1y-?;4O-t_t-bzX4yN4Kto}_kF0F4o#1IqTAvW-5XKa86Q zyf*AQ7@WM|=m{)BdE>D`n#t77Z_5gxULj;~wUxuE*q4!xZS=>H)l%PA3dBvCzc<35 z&~CsMVEE1LWXF&6Yvc8oAP7hvzpnI12tpCA@lgIqie<*lDE zXMad>@k~6E8zA=;iM6gZCv{)0TJ*9vUC@lr`7qRauFtD^(qb$p8FsW$8cJi+a|(s? z*ag~T`m^odb(gZgswy$sRp68Vu2ZiA|~u48dve; zVN;BBx1w~>t3Pm>e55dVG81F8=Xg&x?J>mL5T(pp67A()0Nm$Njp*n^8i(2Ky?>@Z z2WRlGHzTHjN+eJ5f$HDgo_)kVTQwQ{IirhK?9ywE_`4m`yJ-XbIk0h2(RA%4T4*j0 z$PV9P-ApX~%>9%roK+5pl+7Sq+Bf z+i9w)B68#3K{@MLX$;5TReyG>hR8V}gh3BJ=-C-F)$MUi+k-t*Uw;HW-PEUBEzSS@ zpd-iowk$vQvYyQwH|i2NVLlm~^rs~?J{u@&88@+O6(omQ6qjJ%8X0Q7?2~c)}0E z5WvFLphJWTo&&gKqV#Pv|5vQcGYj`S}s^)qm3?R!4}?DEOde3 zZHwNi1^12vx|+v{R?A#F0vU2?XEX_}!C{)~{hi^mBCMwDdqlzd5ok)gzT=y%*1$NR zE3LfLQMD^La9?Kne1C+RhL6DQm_je{(zL`4g%Amqt^Mn9_-p|=^@Z)_A^-I+8kJ_n zI+_LyR}7e1Y7PXRx;2M@<7X;KCA7sy{7@AS{G}CHGurww5GqvglVJ}@ToQD@5c?#7 zP*A5nr=3`8>M+cS&j_T~=b~EN>~}SMYUe;l3Mih)@N|Q)7k@q*TuNDbd=j|_aXlw3 zS|Zbf`r?EtYz=6%3=|T1Jr>Vth|EKx^!u7c4zg+A&hG$CK(fE9CT1LH!FU;3qBqmS zk7~h+NK9VjhO$5BSRyXaR?DHPpvrpG^@sd$3-4Qi(1cn%+k=HfaME8mwkp|!9@Kot`|NCAbTqf+A_ngj6}F zt(`8{E}mXW@PbT!hwWM1*5L4m1vHatEvQd&##~|8G|IlLci`g}Hz0-0ck{@=?W8~~NM5HD#~H&# za!Ur3&4%8tw(4M5iE83$+^OcTmi$z0j*fzdOyf;d`7wV&Y??DNR8mk{VU4Ki=2}Y$ z=N5N#p|H@|0DA;01Wak;_>a({&8r7EdZjc}4uy5w*Q|I2Qz;+mbXTs-hM_BtUxP>x zWSCUhHg`Kqqz=ma;{queMf|dW5-yiacLUwnQmBqRaWI*^)Idau^iSCw` zP;q=(p3XIf*I0Sfev#q$_%?eYVt;9mcx$q%Xuv3%%03^#se{8TADWEdpmlf4Hfsm& zS>&40JxkIsnsnrU9n*$St%%M6I}Xyp9n+MUY8`*s%2bPLaV@LvEiy?;v$m{3&0`%~ z*-^M_ZvU{oxVU>jz)Yfm8KF~Gj;{V8QRdF^xXehkiGS0^HQ1K;>NL6unUvYF3X%iT z^}PqfEGV&YER$h))`f9mGC@4=^Jza5o)o)@k7#qS0o3Qs`N~@I_*^>v{9>tqHFv?n z*Z_akRKh)j;@9g>*_CY&vw7J*Y$=hU?0ic14dzlWQIqmDv1}3+hpGU<#criyqrLgh z5|3QgmJD->lv^-LciRo6(Dughe>1t*u^PoECb1n z#tw`$@o3H!o}=<`;)d<#Sg#@jgLzlxr2-15PIpSmM8cs47u|5H$cu#0S(;7U2zlfElHn7{rT^`vFu7fF9Abp4 z!KA3+Wh7pSc8$(bs9L^)nk!0ide8K@=8Ta33H+2PkT8*&zavF9hUHsc>L)X!+-9l~ z>0I%TF)y+6T?dzDU~{g0bUk`SViSL{C*SU=!5mfN{3H7o)ikL?Hv|a&oN)8I(RJ}D zUVv5k8`>bfbhI-Lr8l5#U!4O1(S7|<+Z8{8nG)%VM%q!U&=~QSVkPwJEm1RLhwpl^ zhn#u7ng34_%wGCLSO*GMXcO(|vk}+7{{@$aRBeVOutWKb-#vdSh4s_Q zy^7HSibkb6f_x1^e$#2aFGfz)r`K>H-)FBBGqX zzp2RIZYQ))DbdzJzAJ_-O7Ez)@FRNTJ)vbznX?LLhA8+8OBH`xM}B|)^LAu`ZHlD$ zKKCxso9|2qw2dePxC4k(x^WWJ{Cc7>lzK~aK0FX&_A<8225xT;AVg)Q1LE=>rICP+ z2jF!87Vwk3z;~?aRW<#aJiTX#UXME7`^*^vj^t%{Kj7DBKTt+6q>T0s;!*`iN(0@c z#DeUZWmK}o^;cc!WHEntKWGF;wE=7;{~wEJSljjxJr!SdRl7m7awT_qSs_)mI6+L3 z!oDkw^vLo58VpgHIE9)<=FTtgyZ$BfF~QLDj{}H{Dy}6bu>N+a>YSW)Wg4^TCeLKv ztEXsE#e(X|85_f^b8Yd`TaBl)V|265ZWYanBiL#hQdUjNPVqR*Fq%*eWDXCx{6^jgjnr?&ceqz~W{f1p!?>86dh z$}}u3xwhl~tox|N%2vd+WcR_uh6++j8^0*gL-X>@H&uT;yJsnscFoOf3eqtrdcrh_ zF7EkOqxor+X0|>wG5l>v4^bX&)2zj%!xoN}^w>)Sti%I2>Gb&uw#yH+MNT_R!@0g_ zJM|cg_2t}sw&eEfHsS!bMrKzz3hJYL#D6UR7 ztOc*)Btd^@G(jfZl<6j%U>MlV6IKzrlxSfI$sLGyK$~zqavO{NrlP@Whg%Pwj{k*X zD+%k{-*@8*txdwpDD2gmW11H{7DcaSR^)XwtSAkH6<8;%P{R>hv_9|$XrvlI{mM|| zm%Tzj3lKNdRoIC86yZ`5W)qaB$}p6Mf4Z1=KFfbo1ES5Y)V^k_poJ_It9EcFVrpe3 zzajI^1JV=Q&qaZWJcA==%>rW}Y=`cD%F!^&wCm-~fT;`9$_$tBlU|Dzi#G&2Q}P3N zhAt3P8cgt|84`@85A3c)2h0L;y*@Dgc^M zK)ipf8iJ7^>6Pp^^%%ATur?SdI8)^kbOKnnBQmuP1*TgD6;f>=`I2@y>TwZ(8b*`B z+-1R5=Xq1ESG7E>Iut8E|G#Feu!@>6$NT@czik&vfF;*Sco`3q#QIUa@xV=%TuP}1 z*3PsE3eBDbYxfZ&J5J9BR=~!@a}Z&S|Du0kZ(?lSF*tN73g&CXd!sWD6~tx^_+{hV zQu#JJ#2+&VV;>(}M7B^T6goIF&_<*ihUkZ!AYm*4ZHX5;p9SgCI}Tk2l2F(xR|dIQ zYZIe3j+M#emC-{2@i|koo}H(P^NYQ6>be1NP@oVY_ao2vPsMa=PLNS!=^J=?F<5^| zhXhcwXr6T4` zR6hPt(!x$(!?(g*h~(N6crK~T3$rKN9OjRzFNqNENk|q4OexE?40L(y(>Ux3c-U=s z7gB^I214WED@xs3_~H$F`5GaVIZS`&2m@+pK|Wii$ECe9v;lHKXhLye%heVQ(%*s| z^M?OAsn?PMN!9mMz~!>QXzewWjPL_8uN&YXvhc>i0m|KtZ6Pl!1|Yg@DZJUeOarg` zO=w1K+NLIgo`1Qva#Uo5hu1YH_nCgfd{=PODrzt^9~_1Q@znTgke}?JxDJ2y&82+1 zPzEm}Ol+5^{*{i^_JErHGZ+c!!`s96N{-AC%6A|WC+V6rcFiJ zeAwYrAh&EyV*0jj+Q{Ivu9JUk!&7(YjiU^}p=24_LMQvipZ+p;os&2>?m%RuSz5hm zeEH@px{KfBySX4?0JK^y&MiU%gh=Khu>r5~gRZ<-#d{RWT5s$l;T zp>*X;&%8B6%j{A^QiLeZar)QSG0wmj8@?r|uLsm2c%f=BCytPI+mnBBeB`Y*vL!-Y ztj2b~Kw+tBW7=BSwvhE>t6AInaoH~KZZmSUgP;OL8at7%uoiKZ^eqIY=gX29Hc^cx z$jd)rgyuHW6l5q4QJ55k7oN*qF#mf&M)U+n%jkU#WQx2%sT1Q9;?lg9-4M^g?UDrz z+wRe|iEw)6BwT6xd4zwPDmab7MfV;$x6Ddeo3A2g{Qup|2?+!JAK}lcS!#8!RJxba zRwD_=R9SV~0GIqWA)zO3u+UDIF^5IPOpuaJIdl897FJX$0~dXNnL)R{jTBeN^{Xwb z2EzgZF4ZA%;+u*(FVBab3_}ScGoaklNRnd8NRav-?gCDsM1+5Yn%fu1%$j@71uYf? z0jxXGySG62nu6@GKHq(LHE6bPEW09$&zA{yI;hY zcqGehdJ4>(QWwwXoomWelWJSMm$D)(vuyO-o)m5If-h2sX{&_Iv4v+&>v93!Zw{f@ z>|ympy%8Mr${ zQ`~Waim(@%5IaDNC%?D|+r7bk2X(dFF*M-I_rafhFAz`Y0t`mIT(*?*UC?%{aiu#wb8YH5K3;#H%Ms;8!MqIv|Cw zgs5v8x)^^;P}bX%@{)4ejlg4VAz17Ex9DK}2GkkidEP)*(U&p=J|dMJq-D)bZr(RT zhq^-HTvIpXqqU(khR{h=(mNVbNb)cwFLH4i0skouimpGpoD~=4KhvQ3PABp`6M|l58oJ-&6)mO#?bhq@JW=Rep#PBFqH)9 z*%#N(=h+Mmf{ zMVdW2^x)Bgcak6)4NUHj=w4%L6-NDiX+@Axcf=AH&kQ1%RXatHP0$PTz~wz`qIK*Q zHG!sZ3w@=LwiN(=Y3`zwW+((|mNGkULLYy5IfCZKX>sWPg$kYUMwlP`kFPbEeY3@} zNL;pP=o}gm%=hl;AQ=B#I+3_kD}=1iKF>*%7y7TE+A!IrDJm2O$w&2klY~pm8i;ro>Z+J(tIgHvjb_MZ1QGwFz9fHp zJra)?*yVrq@_%)Yt(DtZUI+7hu&Fhm!xdJn#6rml(|-voanngbN#@w2SVF``t1>y> zKm`!5asF~W@-7O#@XYfa?|Iuha(HX-nJbkM+AmP4k2&(*lm+-p!k~u8SfGJsk@54!P^IDfFT4&1PKB0~njFnB~$Y7>G*1zae&7 zuwk8JbA4US8_}oC9U_Y91FAu$z4`htZc8OLN@5>h{B!YgT=nn4A~@^iiHLtqma&cD zeb)67X=i36zYhwKicZr`Z2jZ^;&8P+?xfuST4H4!mc0J9P(~X4#GgYq+?bwV%|+oY z^PWWBj8Zbd`*LKi3XOGyKgZzVk ze6tG$JD)uk8?JI-QKQf_Cp`jOzq}kqE3gZu{L>RGww0~v%%?7Kv?Qfv$*$X(@XBS% za?_gy6HZdE<=AR?pid5i8mDq z=^O%f5n4><13t{8F3E>v%ADBSIv$lmVUHctu&MkGr{km3li_zfyUf`z7OK2#K($CJXv%Cis6OidRxMjIAT90$ z`wPSjvYqG0ln!z71?GP~EM#a8uM!Nt{Kv$P@w|2(OKW``fjs?*7fl>1mqX#4uJy3W z9|yh`xU0qh-&5@pKGTEu7%y_W6m=&Y?EsJ+gUHrbgD)RV4FR`q&`_*F-nWmG?q^I(eG$RA`RDv2OuK|$c{qOz@Ab%FRjWQQ5L|m*2Vc0!Uxj_xb2dC$M5ZyVH0! zV(umnb0XU3c1>wTvL+~mf`}J*yBm#-FIG&BcCngN@v$W^j&C9O`gXJ?; zMi6by+ieT^cmbQg%-t8k5bB(WM!Ns9Z;5&GA3f7lK4ti5#?8I^GQuRn8attm)l8o6 z2NK&$@R&ql0Tn}aD}GUn#^3Hosabf0UDk{c6wv0?<}825yBqZ<_6f0>WepK!LIkw~ zxf1U9uMJxiVmlEPGQ&nl3*iv-HBD11R=j%xwj_=EiXF?(z5Vib!0`P^sEw=Ydn=zXt<&)Ww_&e^nB<^G-8XZ0MK}YAzMcmv8|>u)JPv z1ZEwHs%L+29NF{_YDwjKZA|)$jiYa4UBdB{<%>`;U`eNp2dk{5N$cxz(`=o zBh%*_L+8)~v(oj-UwaodOHojd)*Uo=v`ww^>c|(BwLf!DIwnz$&rBYmUT3^*A1SgA&}o-ziDhq$ivk)b@CL(c&g#5aYSwn_Yym z5%8B`L3PF6G@cx9g{64`t6q1$)lNgXePNf90*Ui*3GcWob{W z$bEH)1kZB~sfl6=d7V^sNI}Wni6f|Ms~2@cg7#pUK}2PuvAj>dMqvkH;}U0X`+Ae# zy|H7Pn3&e%{a$Fuj6d>LFW}t3Nw|M7AiM0EP`)9|6~}_)kFi$ywz?1LoVL06qOO07 zxrL!qQY3PBBQbW7+-h!1WveR>Ju;lQs4k7% z*AG5UWo4HfMHQ!Q@{!NX##ohr2s-;bDPioUsMy_`Dwm&0aq83hn$U@-pNRBcV5V!) z9|diT?l#kr63$}FLs`j9fY*PydDP^}M@!KGON^_G%cFWvVVfBsjd~mgs@ic^z;<_) zz5Jd2oE+}pTs*lXIl|WK^BUWLh-x~qTPtRmfVKqzo9 zZG(X>DSkRebuGn!TQUbZ^o~UNS3j}*v6~mYVJ@(I>eXfsBpOVLkXydtZetlZg{qq7 zY6aL6+VlqE;-GMJUdh?IurnqSOejD4@m@!zz{Ll!K1(w4!-f7?vt+|qbi|5fhhdY~ z;biZMkjz4Faa!_tGPQq>YavnBsEqFh?Bv+bL^Fg+FM!#!{)%LPLJBPi)vH08x@h~g z7HKH)g5Dz{l)kqpXe9|QY2H55!O0I+WIk`ZNOE`4%5$Z@x~$sV49~Ij2ZNea=?Hy2 ztAQ~M^gs6@nwvlp<32Oh3Zp1CA3at-d9fHi)8u%meM&UG%%XpD+qw%Lyb^XuV9Yd9 z&|4eZYL7oVL^nh{M|@0%8jyTkW~vuC`YJ40#6xO#@i5V)|8%-UVOkS)d9hIhQ4IR2 z9G9z3(!dn7HbIW)<=i;22f4I0zD{j4BF|A*t9~FM`w=APySv;f9Cz%c7q8c5a`Ayf zD4{LhSLq9Yi$Z_&{P6U_lG_Ka9O?#W?<&m~OtrYQx-r^|ncP`uO(?8T zl_=``Pf;IfB(rAcxGqN@Ce#p2dOqE`EIF}6b9+Q1s`G!-4fcDQPlb>>kJeo@{(p-5 zIna)I*OhXJBCZ_#*m$orCbK2!2{6|m>G~HA(w7yuIgx$=>3#?aN|#0aEOD_6MWrON zcQ(|Rjgb;xDr9+UvR7U83Io^l?Zlyl73tkQgN}*qw^FT#aHkM9gS?|uirx>Dzgh{{ zQ3Z&G!-0QtVR={0%2C1GeaF5?c&l4)2p|XDO0jchaNj9XMP{#{d!oBR7-=-q$SRBF z1Kl%Qd&fHBIr^`uwYBBTk>2{OefiQfz{ZK|KwP?9Agxu}&23IZmo2t0 zo~sY6Eht>T3aVj}3yEiKwWJUg255@3y-7pw$~qAiWFQJYqZ9glGMrkA z!y0GBh(Co2dBdIfKm~8z>o3karGH>jevEW@N8O{%^_TtAU&G;xYli+v8=+XT?2ucR z{Ldkc1BzLYinO5ZGLYI~uS?9IdxgI{{{Mf395uh}FtoFGt+SrIZ$e`aczL6mDcIyP zMDzqA(J9}kATXZmKT!1}L!ur?#_#&q&H>2iP{J5kLX2`Tj)AqBe+Zc-1jIXkm2}c$ zF0vzhzJq7JYFk^7t(vx7Lph^2HLX1$^}U9D)SG8kkj4n?W^M;Ninab*==HOZ9(8|Y zLtW?0_C4Fvx(Sp-`c@!hc&R~C(IP8K%Dj(KlM0Fcl?)67gqe5FZ@F6x7>o6sivFFi z{U#3>Q96Xbo%GqVXLj!N>(TeLYcj&bZ(!$F> zvnJiGhqbZ}me!%y5kj(yfUPGOma<=GYR+W(_3QR}bL{p5pXNsU7sVNUnVsGs#$8FD z4)6DxZYdCWqY858|J_L2rf`4l5WGZmU;A{ElO(V*bLpHsSEXwWN&FaNuv0#(v`RC3 z=aay!S-N5O_8|Fx2SdnAUTKde@7A(xMgIeZd|v)_<*xJj=t{uB^%1!hf6xXSnV}fa zS2F|+-1_MWbdVD{9RpLmt^;yznNc%2r?!lLVlK|xl$@iSS7WlPHJ#+DDm6N%4}j(>;o|CyPD0$ZqWW#AvP ze)l!h2~eV%eK9-Un$6UlDrqI$4Q zhoG@~&#M@*k1GcQrpiF&@`R;hCc!)k9stCOP>{}jw89IIf-YYr%CvxJEM$Xp%qmQd zNb`)4wt8b_jK1r0ztTX7V07BN6G{E3vcsKr&f7_8{OzWM6sdoi*d2%F73IP+(nw@8 z`!}chME9T^YTUUFiQBxTBT*j1648#Uh~r5js#6N6}2o`A2e#`l`7nZvLIh zp+BdliQU%CF_lr>(sRL`L{Av$;w^P^gN^`#)GZbx}yo^2-W|WqAaoeQv=0SC~yy zU!HgX6uBZjIl@wAMH?!ALvmi;H`qT7WyFA%I3y0wRE>X0jv2IlAnvn96&x{2^d|m0 z#c|eEI9Y{iS1oNAhyf>yw(ov@vjq$-_RetAaL-3qz+E{FGW)md>Od;7Mr?&YYc+?j zDMfF$$zT`)#e1Z)U^k)aoJcWJYpRJ;czq7YnS14lN|Alv70s0Q5at^Gp1lVec-;os zuQb$vLRWu^o%D(AD^9_dY$49TtB z;wMET1wVU81bibea++)vl~WtWxjd?iCJkxSk?`G`tz^JE>OzlS0}5bRyT?;s&S9zk zIjhxs@_NWPO~Fe_xV=Vc0W8qp+TZ3AnJfb8UG9G@`8{jlwAVnzVw%W!Ell zcWkvN`ygWJ@#cU2zYB@~ULj%*Q`GEBP_b!EfaF0bJYUkdc5QyT>uDeWxY;TN%D&Ke z_`KHp^4b?WoKizq;*yya{0@2H=3wk~0-}8MYzkVQ?J*Ht7~Ve9^Ps*S;ihx7?jHh$ zKgoX*Z>)p@w6%WSkl_26M)cK+diTJ!e_B0eW<%d9W72YX*zH)XX6J?b+!)C($*~! z=^)TP0EJK7kxKdD6Ye?x0y5}0aGQl~S;GAeRN(y3$VL_4R9fB2n)L?j8Nktv@6|mH zFp1D~J5b-tUrj;t)bJn4tz8wKOE!9d9~~D(YMd7zL{=jTcU!#(3PtK8R|a4ewGMwZ zT5~?4N(LQYGU%2G=0fOV&!0(H+X+sq1q#C z@-4LBCS!f)Jc=W8qmO06Q|D_eh-1n$jXbCc?R&@OL_@L*m<&J^3xx4lnE?my@C4Bgo6}+y}uU9Nr@V;e*CrH658#?=E5B z0RtpiFiB{Tl?F_D~iN6|0IpYhCe#_X@LZ;UMsY zzwu4;A*t03j~Awovi-`V11lp_dXUuFOpb;m36Agh>ToWhZO6=I{;xBh#6mYwfiDyFaX$#%p0d#! zm1wXBe?&!{w@IN4)O~d5V?<{0TdF8)m(wd7yb#bK{6TcMzpmG#+;QwI!0W5OWxbO;j;LT5z?`N} zrHQ>?U=dN+{8D?;E=(aue!cmrnRc0Bj3YgRtp`#+B86#_I=Oy+ z;y`gg@*scgltCxQ@Nor>6@)AQqy0(0s=U~mHG;pE4mW-?Jp(W@TW$I4sB)UWQnfxw z0Jq~7VHoYo!V>V2FO_I*RWPkZU&(TL`#h(Mh(Xf{BJtVU!GTT~!Z%bS%es2|m(1Z4 zX~qKdBvN+|jAPPNGX(3gJZBR*j{xoYBzvJ1-~xXFCd!oz(vf?8oPn7Q9J1bL{ULI?V&9__#0U?L9utmlf%BTk}*%Drg>8br0iupiG)1IjFXX9$s&bUm_BCGJQRaqf z-dTUdqkT(Ad~KV$xMv|@@x0M$=8sA@wd!0oowf*K!_O!I?T4?kf*5xHh5tKQxjP*UA=J@ zW6^U>s}RoMV{A%mZzg?1f#iNfBf%(kGB6fdq z-DAcW6V(03r(Nuue~zJv0$$+20Zd@XJ1~Ns@h7AOt;W!gyS-}qmfQRWx;c3$fE@js>mU;`v*;)2Q$Y+YXyuDWR2>yd%vi2>P?8wD0kkwBZE0gY^gKXFXyCf)p?c9`%!CkG?l_r8eRTi?PjX>e=)Rt6iI!Je zk*V2W^q9Y_9&RSde(VCF=Io<;9(O-e7TufYV+W=Hv}VZ zh}qz$lGGMefGwUg#v^9OOyPvtOEyfP`iS(#&o^{jp~Qm-ag?uMiEb(nt-iNqkVzpOE; zk2tL#9&RBfA02+e6gq{HUBnNj^Joa;v|p)S5|G<91=MCY9@0=VMlXDiT|f@$m(_(e zp0Mh2?=Y!z&R-WAPt3DH&uHYBttT>2 zCT~Y|aI~_reFxt1G_!x&c!!}=5g9d_Xl2_eilBZI;lfCGPdXADcNf81^Nd%K@+cl0U#EY@BIozun#YbnU|0*g>Vpx>O0^X6mK&L`pSb{DT zo~HT!c8B3Na?azEsm5W$d`r7SHV$$OFH(r$Os67fc#Wrb*x!Ge=u~-*rj@`&0)q`BEY>KEvEmX^{OBYbyfjEpd?YS9M9Fn1QD{X7gC?&go3z9tE2%RYJ{C>fE~A$!RH-R^p@6h!1dh4 zyF!&u<{E!fE|Oou-lZa!;AaJ>=dV)tA8VJm{^_LeAXEPZa@OdpzqaGGf$JGJ36&)N z5k-M?B=;2b{jyr%D>ghY;Vul^A=pKQ8jFrnaZ;{qW3tJib)X#03w(z~8;z6#E0TSRwKV^+S=B5;W5QiMn zu66b2$H>@-Rj!0H-LB z1x&<7iv`xd+eEYP;BsR2+=p5`w}PykPELPt2#3B5V8|$K=A_xsVAfmfZdjX7p(qJ_ zUwvi}=Kj|HT#n@V3P2xq3JxB2jSCYApNX5aP&6Ow*@0@~Jg6 zMC-h?|O9k%I3CXM-^_&o3K6uc{9>6^ZXudMdJi4R4MDzoUk@^V@ezxd6>B ztZZl|V!7#xAqSO>%3hjV7wHpqfCz^K6Q{+!3DbCbN@Am3O?tu1uLlHLP^u+QU`?>` zGx!?8j7IG4XbWs=tG*}k`810JwKacMG@euqRi-PmS7qUmw=`EXx>}N9j%HPSMX&FyQw|=1X^f=1#NUx6oBrkJ zw)9Ui;%R?VB@wzQy>v|f^)<~iHLM_*SYrW$LI*qWXwBT47L0#YrOJ&UvJroaAFB>8 z`#QH1pJL^yH`M!dklaHx3{+yt=Vtt; zRCRiiyb%Q4x@*?pL(TclB!CJl-?2D(R*A*%y^uQYu3UT(p{_g*^@6(6OL99avTZ`& zokdm>syxc8-_4k#o3__|c}?oN>TuYzNGU2Is0 z*0x+_GWFZfyxz8GhNNKjK|i?*Wme{5{4J3?&>rvJPd>TLte;>*JcuyXzk3q&)a4Yx zn7(+nchg-2QbF)pWs19sq9-q>c0mi{zuuWO|JPAqB4nnfBj6m`buNDtVO~Al@pR*s zl~yU#eMrN~IP0TQb3usZ^^SjkTS^+_HjL*dbhfxlG)f-u;@jITs8v(p>Bi*y>Y>jt zaJ%>)fJ=br0~ffg^eN5xoGLYLWC4ZD`;F2j3knD98zAgRw-T3}m343b`Ucf#FEF`p z#wQ)=xyN6Jh-%CKY@B}}Q%TI;Jb8{<^ac+;UuI_m(ld-WBAH1S8Ai8KYUsA{EY{~I zZ5Elv?h{`ObU{{(m7{9_w#UF7lVhU!{Gm}x6beYd{W3+Jm!juyNqfChti{{o+$AHS zrw3owUnPgU8hZC%TCeV@#*|9e}}rztKA7Ia`z&o1cU%M6(r4^_;< z1jd;jG#4%7v=u_p1U!I}%O}jB?L*iAvFPiUJ?e$yhgp9&Sr?c#5nc;MEghgoNK#*K zbjmQ|ec(45=^x$TJE;O|eq!MN1eqwGP0l8#SP-2^RVBveCQ4EV?J|-48NMC7$Tzo2 zb=iUUPsKp?8F?>0hX0mZ@(mLmbvc!iF#}uo_wU?pX+m7)L+yb~e4%9xG8uYHWPYB5 z(&8lYs1|<=e*v(0%LP}z$YrL@(^5*_mzrRN{%bG8RQZYvTmjm_5hpYp8~GrL?A+3r zS1gfgRe**0r-9yXWZIZl2Da?QG%n=H+c6Qd?6A{kmqV*yOTAiyPF;%EJE!Oq#g@5} zyE)9zeZbTUj!5?limAhS%efre^Kt1euBUMT!rt8>TdF3$*$q3?y?Y515bh%dF zbE?-Wl36i1l0)y}74fJ%yk{-avOzTeC0R;*BUkd`8B`W8VY^1@x2E|h&*#G8wvRW6H-mUE3-JQ(HFk029mwfd} z;i58dBAzIDu(Qu8SuB{%cNQk-X&B)i6o!9$_3T8gs^z?=!STUf9tQ{r$aE)wxNEoP z)*%D|86Ge@^{bO}@egN29#@4R0UHD^DXwY>9lFZ5r}~|pYdvW#$BL|X3~dmB^}FqSS{-cP!0TwUTKr?y_&eU$4nNu+-LM!SsEHVdX@ zXk|)D=?GL*(!2c9=LZzQ#F)!LK;^*2)soZ}R(7xS7;xEq>F}sF|51Pc>(El+68$WfU*}L1% z)Q8%*sVLCgqMjY+;l6upMjc}(b67wFOGWM5r*`#SoD zM$ex7-Ay~9S>}KW@zve#?s#Ou5QzugWbJK+ltr9YXzkL_e2`*+CYCe>xY~bbMN^I` zpITL2B#l8JN~#?`$RXPYBgq(oscZ_Ea5X9j(HdYe-?!jPWBJoKfG2AUe)BKHFSe%2 z#8gEoKfD(rvGs8h;!56{cxDdQ_tcnPZxwf0zN>VW-)vDMG{yAExVM_S&GQxL>Klkm za37T>x>7okxD{-$o*81pR!x7|)lyn0xKwZDD&Jn2${m2k+%bPvk^;!@iRBOGg1+9G z&2dG#we4;)iJv)#N?K^;b(5C}p%AtP@DZw@dZ~$^_m|KP-2ah>zo*uXcZyqHdeq88p;rDr0_x z#^fh^we@qJ64nerw#%iH4B8)kuBk<@Qps%f9s>1#7ce@`=$)d(T9Lj}ajE$yJ?Zv8 zpRB(*owmdUHcLgpWoLoaNt2fb?`k%4XiC>Aw$LmemFLW^K%tzgPO(-O z^B42XHp;OJOdPudV^Ydy8+%p17}k zOqAE_@k-k&pQLw=HB^w!_cM?kqYq}{?*o!sQ5XCc@FG8SVe+KIgBI2MB zNwy3%iywaiEB0N_gSd0!@)FA|ffuk^v_?>jx%ovHBH6-!cSUb%-A~5AWdT}y=h3i2 z*wIZTP9ACOiNC;gLLwy6%QxE?45((ZRr2nzn6)vhlLw%-<}fU+&!a3d5#5^Bv}T<)9)}BHVw)iHVwKh4jL-rZN|_fS55zm|fA( zqODm@Gfxn+>k`Fg0)K_-K+e_WAW&u3H+RGX{XicLk%=_*wJ8)Tjc&ieP5FPLH`=Z; ziPxK{OBzKYsMs{J^*;yB@p#^&#w_*WqzAWuQ5>bu<3%R-E;UT|b;m(YDot)5hLh?H z=ZAkgHNyfHWWSe`osk1uX|Eb(u5MPsXn|Yq9#2Dhsl5YU-~ww5nwR8KFNGxt7TeHM zadzwj>xwiZX#p`xIFzEqRjE;3VqI)_XI4W@URmvZmzl{wEt`euaY}BK&0il6$!o*< zq;*fu;>4UZUHwoEz|I)ffGJTRCd<6mC*Obgh&~k(nWmByMWl05`u>f_>YCFluAN~I zoPB+_zX$Fm78-)#&@cJh#?dp`%j~J@ml=I=%pK*JNKBT;>4q7_7 zMH%c<0%1bT@?k4Df=L#Cs2j$*U0XJjb|@3z9o6xK#;Wso%0D)d`<6iwoU<#Wl^Y|R zklyf6%-=Jj&v7Gl*5&b!Fay__g%`$?M5HUhVT*kipRfTC!jd$Ztr}(q*??}}?yVxu zDTt?YN?~D>UMO3TUF5w{?hK{14}yQ?ww#jlY3xQN$MC_6!bl$@q94jOeKuYr-mB!{ z^`)lJjt>4*z-|^51_S)-ZeCWCf&%rb#-mc%zIZ;P2(w)fKjM>>y_4W$&| zry07(BAL9K(5xRr0kQ@i()dq60ub7bMBN`-K*)p4&Nxqe%&ruW zE9^{PPQ{e&P`wSDwwewMdOrL4gTW{<8<^dJ4NzzB!JLV>pzbngcmrX7;R=*eC#U3S zMPd<&yZ@mxeb$jUuJH4wo+5iLgTP#30SBx~+PL}9Dd=qT<$sgDH~w;7Z71aZTNW?t z+4NWEWa@xNQQxxxNx~!{@RTj=m81AgRj)96A7{h2i>NZ<2?%z@#7MiV{uL~m(9_tv zNa~R>#7*{nW0}JY6ke}?gmhVRZ!B@Vx(o=4eC~C$me?JxG2`=X)Dy{AZW(!-MP%&a z)zhtj;KOASljp&OE!WmauZv1)h2=fs4@%B$zGL<=FXLcLL`VvC8;s!wVj${z9yG{X z$FXtRw(tARo%PZ~)wPg%7Ct}SE-h}nTH9vh!nfKUF#^D~G(Qh_FXXMjRBHSc9mC-Y8ZD$_l2v$lUIu9BEYcS$ zf=iQ$h2hC!1p;AXeFM+996?WX#ApLIor~ zUN?;=1T?!ij`#qt2U){gAEgVEy8VJyD4%#2&R4Vz+b=+P#jCP^=@z|0O}xTZ+M@^> zair6ZJD9Bt17(pakE!odPfa;SdK^@;FIL!jF)8a7sAY>@k8K|NL7>Z&$qCmuW5fQ! z-c$S1vCv9?*rl97QMl+i`F@GE%zvc%+J2U$QQ|ZyVTh^v zz><1~y6#jCIbQBhU4y7$~MT*a;lg=*|-5To~6>j6yG$^eB zz^mc1C5^ClHT&$P)<_s(E~uoTK9X$<*%Iuf3w5hkrQEPQLD*>f+=YDsT7q=4lrXJ} zwXO%74q-~J4r!G*`LI-bCt5wms^(r-CXgC#lb$3^PVF>r`0(16Rn`3n zQ#8{Eb}Fsgy+Ok*gF_c;`u|mp-@vwib7A8I!sDS);L~-@tNxoXjn@}`){;%;=?b{k zj%=(~b30?bvbJ=D5&xeA<-a}Zo;T2)*LMz#eG4UYVwexw@GiC~zC3jBB_wg{D&bIb z(iBb0ql$(d!nJ6jmqGQcjgJ7KE-H-fI>#jG?Qit9lAP()%1|9Y?@jjF^r*FeZ6HC# zS@mkyXv6H7CQ0vil5=KGDHMGGN8Tw(I8y>>!-Lc0U0oM|@c!GA@&H1jd}dskzxp^1 z`%pB28A0Cet1?X(u;h;AxS?zq)1B6wdpVjwc%|FOc5@I10FyS^ewtjUXWJ4^9~Z&7 zYVw;P<}Hicq@DfoB=vU__JlTn$#qAyaKL+5=PVd9zunGV+Apg1uUR9RbVz`|m|dEX zl+-VT0cHz*?9!zrH+BuTm?m=O>0;e04o&ccxbZ`Yd=MmWkmYuD{bsVNIHgm-DAU%{ z*%J7& zCOZs+DP5_puY{BHCx*sBQHEVjD!2^aDrK({kNKz+AgVMhwq&v|R?X3<`(;~Z$47n> zyGe`V#Wv3uj37k-X5!g@@gN|h-YF7c zV(SXs6pISnKY||#LdOl}UKE~AO2S18XR!X!W%M|EwBwZcCU(;(SRKx8qoU(-6B?y1UcLwDV~kfDi{;~2}AUH86Hb& zIwxe`W4FKhxyCkKG+7$0>x5)D9K^jMArOmoJV0G5fpzWOi2%u z#K8E+C=>H!1Sh>MHDpt4(cJZ zu6g}Gh}%&YQkCHaD|F!vL|D}nzw@k50b6Ha1yNi5bMA2NnTc}?f!^u*{IxlXax5VA zuZ7)zf(|6XUcS(@);ROOY+8kSZ&F+{KX5V_VST`>+10h{48bVI__M^1z#EG_0gfcs zWBG~)X8>LEXtu9S380WM8Vb*z5%eG=d2D8TP%miX)pl_hMw20tT)%G@)YW=_B#P(~ z%pOsPF<#bn?rJy>&U7MG%7jXBj1Ic+Xa5m@3IT#AFE@5|RgDfe^E|P;M*ShX8s1@p zQ7QbgO^KM>;!0YC=sQa9Gg2|dQGCN>3eUYQ zet5ejVV5{41~Uh3p6^;{c+3oCr?5> z{X4;2{j{lRrw1yr?uE^(#@{IIY9B6wzI&xn)1Z8B5XfDW5mJQ~P`I;Xudo*gVj4ukg&@j0NS!Rs1D1{dBWVW;_h(0 zh)=`qVgx_)I>Go8I}hJ~ERXx0jM!o_Z3d3-BvLTwq%*b8e*DwDtc72kJihC%ki}u3 zE#d!?B@8!|Qw>I(+0P464(sHZv`@% z&05~_EVM@y)KxuTxrIqGzMRj3&?#S<@3sYWu@Z??9+6WSvK5^~Z!gSXLc&__kJCD) z4aMiO0RT+dxp-!O7CT#o9k*zn=O>@2OKv5eYm!gkQAV0~2`L;P9DA3{jtjlfKkv{+ zi%M>${|x@T-AFYtO=$ooyS~BI1@MC*rU}b3vt8L^BM?yLJ*{~M>8tEI&HkoZp@7d> zTj1|7uK1ri!U4{TdBw7(F?|8M1V{j8)fETZ#4d3~t#t~2q@06uo_R83(lI#UGqyqY z4=N0V3CpL>lBzQqXoYP7Ju%xHy4Syu7BS6LU?WiNua}$|f9fRBo8J079Xk=I2Od=A zfc6Oo0=qfvbefrz4PSP6RdG9N?n@E?*8bZK@W2KVyBkFsxK`K?l(G_+t%Yw`X-~8^wSh<;$*#S|AmZ2qn-rG;gq-drY$dLs^>f0)$rIz? z*24fJ731n>eXgpv{lwydpL?Q#d!yX!;9QF^bSu^#?sj(pz|vA}S59NF<&oSB8Sqi} za`W&vcFIn9K3|BYvO1pE&Q9gN#;o|dQJEMq#h+(?R(yh?R=9jPTzLqT0Z?NtK%I*^ z31!qg#?soWO`9G>QdT`(o4k;7+r5zXdBQ)N(?@bZplgdQJ_ynCDsw39Q97QAnfe@RfhD5%ugW@qLob5;^6J)#P|iji)wh(lNuY&$sCN5QPh zXWp61UTgs5I|q-?G)MCYwF?sY!c^l0 zI2{=`)Zfkf-uJ-O+x-QCn#Qyb=?{~kzOPk(V2D#daZ}Pz)#2}ZzgE_`)sxzGtF;LZ z<%c;7$`Uo&J8l7^S}RC9?r23l8`HrBBP8Yxa8Kuup`uE&+YM~D*htv-j9>Q$Sn;FEs7DAx;Kz>9FcZ# zYB5;JR}>d=Wq_YYs^#t~aq*0x!5Py~GNVPmM^C`DXoT8;`whwJcR#84$4&h%h8pg5 z@u8#(S)R2I-jY8%w4ttYb1IsOX0_da7<`HSnMb6rZi8JCtZAC-f-V<6U$qG`pDqb}us()-_D4dw< zzpSaxo0;;#TVOxvQ+qdjFXb(RT|w$33z;}tH1P(Skk_zYiLB_vIg|)U?@|PR53tUO zc!c4TbHI%5p<`w?@HT*SNFP3EGKWGODqF)r=Ly9P+$vA#$Z-ullo5C6mcbByrw%g0 zHf!iwZHKMl7nfX$BKIPKCQ^hx)JF-#%iTbRyM>WG`iPJ%qu9ujA!^awwj*HZH(P7` zs(+MhfBmRnI9&I@SWWSpE|TznuE&x;v6Va8PLWnK>vdfzGuU|F4T2%aE?x&R^_eFo zV`BShWvTTb`m7VV=zc|%>qm%?0A2rt@1dkGnfqhL#@-rR$7pbkX|~$r#OTApk9rp* zDK;fTZ)oPv1=ECxSR3bB`-PFGlHG?UPh2$~uuIJqjOkjN2p$$+_4_=2z8zvg6x>u3F_ z+ytUSilxLYUnP;B7Tc+R6$oyW5>9=!%-C=M0W%BSN!^O!!?5u(BF-vZ6V_E>VJgA8 zU)N}Mom~wbg$5-6Gh>KjdYIpvz^AddzFn;{5UP#dxj$?AL4|*Vt+hp1{`~ufnX2h~ z<4_S6XHjL-T!4Rku}$yUH&NW<*v52!%3ZHwwpuMS{{ne-p)QYq<=Py4-w`xqUHcMI zo(&<;Ku?wb!&dB2#+VObu6amohtX85z`R=htFxzeXwd{I&l}vX3!Z=mP#^lc^8&ms zeVWJj5dT%z#UNQ^;>&_8Tg1@}Y`3?$>HLkOvyc}EZ6E!~>yJLyZd84%LtE^eOcbrP zUNZ(7&Q97>4~iXsw_GLJ$Ae_AR8+Fm^(3o#ew-x^eeV)`=~~21dEXU^n&HMeV^O81 zDXERN(;fddR-vnWNof$#eRjF#J(n1Kzk=W}6momlGFK2BoF_Xo#5KmMi2kvZoL#8f zNJUZ3fJE%XQw zJl{Mn*#u~sK7MW3>RvUt5B~x5I~);K5zG?#H%-u=U3$ZES~H!9vk&!+ABXcDu<{;% zzbVMlB$6CZK*sNA*2{+9-J%DDY)?AFVQpC*2Z|PNWyupQb*OAggwmLSu5Q&&zjwuj zzI}HeG1YW`+A+^Vu2xZt-7u$A{Y|nTM6U8J@#GmQ5pMyt?h4qyUrxgsNYzrg-LbM4 zg%_(_Y9|t~($IaN8sHPC3}x4M6u~Sv-pGlp2j@q%W6*KY$8IAy!4QMe z5VZvSjznA0IbZvZ-?TAJYVI^bBennY{H2Ig^U>hmk?t1d$`z+avZ>?oQ5~U3BF5#y zXMxc3o&7=KMxO8SmY9;nRq0^~E_>sR-84BH(!ayysrKn^ddxvn|BK^$7B9pr%ObyY zi-dcBAkV!FP&PFYuHn_yL-^gKlKGU;Y?^7YcdH_K=k%V}zN}_Y%8iK$=v2FnfELqi zIP=R~3wEgn%iIo~`o~G9*t|tXHmgllwEwYnlOB`2n68>(F`^VIffW^1%FQBoA_PH{ zLxNwZKiL!JdA|ZJ?~@0{`B9HayTfpP5w~c61~Og8LuQ8%#Fki(?~D{ZqDx!f)uaeW zP+?;>^WbktVTUf3jZ?*c5y%{c*MtmQMfgi5%vRJmKy~oF7NU$AwpSa+Mz_RX6oL2g z0&k;)DNk|2>nQ^X_(5>Zy^kNmp+4}3DjbRq;;uw2m+c0&1M0_ydr7SZ^B zDWPFYb6*udf6ATnys0i2wxS%giP<72exfqNLu+WkU!hL!95z{RM4MB>w5ElK(x<-mNb)7NimLvGh+7xlojmP0h?jflDL zG^vNv?TI9qK4eoL@bNT;2d|!diMOykz(1U)vP>>5?uw46)>*K1A^5NSXa7t+2m>8j zl`&VX=h{Ergf)rAWke~;JK7!us%wc=Ux%`6DGUj&L$-l}=}yX>xX7q~c-QFc&kKEP zcHD6v>2wBYWd6CTxR4GL-lh5v&#jmhRhN8G9U2JW$*gP^REgH_zuR~|Yb*s7Di{Cf z%D&gwm=Y1U&(gA~c7^2+hacFVPcV+F`g>#G5RzU*d5~y*iVg7SGj;xMjpG)uQQ}?x zJvuiwWVei((=Ou=EKtFJCl9SSp9rV?GPL8U3+kiW^WG}i=pR^`YAO(}u_@Q$9O=CA|*l3Sw zI8B2!4}pfP_Uhv{iOMVpVXj_}xYiSqSPs711EG0Z`=Jm8qOm}KKU!E=s&ybCFLZ{u zyi=4S7mx;!N!`<~?tkYYDmRU)^03}xp{NOwFl-;D#BtX?#vt^2^R}W*T>pq{2TLY; zbXH_aaEO(LPhX$=+8%VVnTNPpbq+uK^@=NB4RpOVxf2gK@~~(y^6h#?u0@(0&(k53 z7JrB)G*oiqyxA^)>Z_=g(7okGzzk3y)D}(g`F%>L0870HwS)Dge`ImcVAtGW4Gnuw zGHD(BDGr9b<(8jJ*6gY&#_50h5kM$P|D0HPH(5^9o?h)LU3OIr;feDYE`KWTCKhN{ z-_n&DM7%zpCq8#z`Y(;0r(p5ks)#pFS3~R;RVYkio&@NBI%gV4G5TMf5d0Swt~1B> zunudAj~m!_cNVBi%u|BDa{%>sKhfUdP(2(@cF&tSf5tO68xGy)tOW;ONQgaQ~Fe!`k zv5-Pyf-mEK{h*_lDds0z%lfjAhN*o`R;h36?Z0j&twTm}sfq;cp8t3Rwcg0qrBn5w zdIQ6$NOd2rPQ~8|SH>DVfOVrOFQ}9`roQW&)f1b4yUidS3ECR#MAp1Wcj#c8?(B*z9_j*OhEVzIuIwsDwPbvw>~ zt1OvUQ0v&=F>c$av^)fB7rE_$vOfRNs-K46hS3#o=5j%776+pwpc z@9}>O4W>C=P-U3Xr8Eh*#DDB><;*sIftdS&)X|pv0nCRIn6XCx`MAUx)?DyPI8G3n z4zX*V^N_}C(%ayoqudCT=0mfUi2dG7Bjqm!4fv) zkKCKnj1dpx@UtuAreYzZ?y!Zb&sx__3oRS=@7)4DGfaNg&_s?>;H?K_-41^V6?>4l z+~soRcgOMc8xIY1#Z?QwQ^cQid_JTXn5vWbBoCk)v@-%e`t+37VI*HSjH$&(F0eFX z1ftw}MOsw5uch{~VNz5$KPk|EA}X(`McQ9pVC>a~V@xUX=F|*ezf}+&?1;wBjMZdI zp;pm+vF!manrXT5UAq)uin@?PGZwWcG_RinKx)0oVif%I04-bE#vq|1_D9Jz6yIy; zZ|T^UQ}(8!QdI?Wo!gCvs^%qv%tO|RJ8q6d&-i;Cv!OhGTAjn6?n_92uCrt~;jyy) z9~8kZ1a8Wrsj2688yIZK^KJ}rDga$N^g27<10^pVrNKK<69EJs@%a=8dT$l(UE|(A zWTD#tZCQ>GoRX7rE3evUNVaexHmVkT5g%56cL#0RRL~9s*=-Y}?jbhPOak3v_FP5R zpT(f5`N+0Lrrkbs+fQ$QI2TJK!`C&|AB}8CF8VRf3(OFSw`BUtTtz(m^_auP{7(@F z4kCc32u$IjW+^nuf1?A9{bkCY7nJiNnQ=5f!Jw6QHF;`x?oL%Gav?PuOuU6VSjy@e z(1<-cW-ax2uSOT;)o)bHJiuJu$0x>Cio9InSNbSx@4X(MLaCu)7e-#yF*zA3}}1+=nQUj|ALhTe{s4UvsKnu?C6#w z^d6VABb(>K?Qg#Yv=XnI+5|}CTu6!2Z_~xP&)}++Po()s_WSN~R&0ZsONsJ}^Vh{u zH?)jNtW??a(8DNL-LXZFs1ZX+u9;L>cyj;%Hr&NT00GOC2a5AM&qfntJ}?c70ssI2 H00CKAq@i)U delta 58421 zcmV(tKEb~9=bWcOu0nI6?GHlV>a8?3j; zowNpjG7#~Mj*5-NUB?vs6pe+@d$ zWlE_$>xY@7t8~=IrmK^m*y%mYV#PGyj{68CIEVnWc)(p7>s7NEh zVrqZ(i?U!OQhaG7ECmut^UGz}vIcKOEU`zAX*62VRcL1Rx|z=9?tr&Je=@_hkd_^{ zg$eFv)Fp^emhn8@;}>lDO8wRIGtrjuBv`P;sgk+&g?UB8JW~rd8qi|NwaNPI{HS{B z--_}y=i`m{fm~7ER=ppsPjoC?pl7>H)S&ZavI3xhwIa-K2i|)U?|Z$3hTpy8Fo#}6 zu-`!U)bQ4UCHa!a6bP-wf7)C(6sr?{buh`3HK9Q6Qk3U&o3OgE*zAp-JtipNkaT_du(~aN&BXZbBea) z_G()6-r^KWHR2gz%q|lP+p^@n@2#24Y0(?ia`PoK6*O0}aZds*f6MnTHu|nIsUUhx zJG4#qo??MR0%)Q+7=Z8S^Ds?*D+;^e&vO+cES5XGY)N-ohylwO=}_4UCWoeiD+|qM1&b)-C*aa%zXI zct(oma}<&rC}#1Af3D((sTuIAXtj3WFAuf39#{<*aIjjqn_wzGtbNlH`*qQm5?hak za>5Qv)n^hH>(UzbYR**P;Ff;xYoiW5jj5>n7p>PC^5J)@sUj$9&M#s85fPO5CV^fl z`O&-<8`X->NS`)#wNw6)ZlVD|PlS4NeC^ne{_V_{0eREhe}{JM&CD~*W#`SE{i}+} z@ORXGrgPZwyd9*j^U4OldMGZ*S-Dxl{Qx{2c}~W#G7?dW7^!)uyjN$yKYY&q!X(>s z>l!kV7{Y6U%M}yh^3v2BUyU|xq4u@CVd9cX^&`^pTggb^KB_fhVk@+SuEgPaRYmIo zywIHyS@L;Sf2de=*d^)J^V0HJ9kB1dHejN3T8XC7oL=U*rTGWa2!*)GH}F0kT=Cx4 z$?mm_(v;(XH91m6FjQgZmvr7o{6aOEjanKNEyOV;P zc~jCL;Q5)@>JWnx*Nu3&#eb=IhYV6_jyE5^_7bJof3D$+^PDlczQl|WvS%y=ky5ny zK*Ok18!sBS^m!BoRUin+R12K5-J}k=oj(C2TPdP*1NqZ-_J@o6kn#wboUt7i#NwFQ>WEHHFRe_kfYFqJb*5~nnElJ_}Ed^hszzmaMT zWXdBg`w^K#xBP98rRZ#G5c{^4Z7UGjbw;^&_N@atAi5H&7G~vYAJ4;p Ep!AEtejdO$2K^q765GaA ze-MKBpsO3QMRwVjg4P${H())eBz3Ij0}s`;^dFl!F3F^;ft2X|vOw%x8Z%(HtOzVA zkSh8h<3n6O?MtHL?!I$3I8+X4vK&i%ZRmm6FSS9;fDv5MROBPZBO2mgDsTQKDMhI% zkn4qKvwd+bS}{P$n?GWHd663ae59w`W~hjlCaDH$x5e+=4} zsm#304D#)xB*`UU4mh@CJA@*k!Ut1Mid1X7*9sLSk^8T?Lfvr_nIy^kf|2!FdBA}& zWfwfU^Q%*fM-Qnm_l~4+Qy30Bhk@>aNHOz7T>LRzAshqGR7iY=Q6e0x7af}Ohqs^e zs_Y7w^dh!9shpyyXyXS6sz5bIfBgjMK4MIBHQx^ZYYJ+MHyu=OI8pEza9siClt_W_ zYL?+x%Ch{v4sob+v!@`r%k=(~Euw`jw}7QJ_nre5p9nkzB{iv?Iy9=^2XWauggT0s ztie*q$<##O92{Q~-!HVMXXyBAG8WOrY8pXmGn9Y521Gt7c5?c_pvtVUe@%!c-j38s zVLG=8;`>`2aUx@y%9?k<%NAuak-3yyBiQr4x}0}p1FjrRMRKM#4rjmZ*O93f;$E$5 zNi|o||9_h{&u$Thu(|#vq=3P6vRl!iBc6)egD2cAB=J~Osaf383zL*1e)wd>}s?F8OK_$#3CQ3vK|2c;Ny$7#bMU{g!u zBM0Ikgul}qdF=dDx8z0uAN>58Y;5cbEsKT&64{CX50|4^mc|1d4sUU|n!#|A$? z*GMBa;RRFLHpgRWuhTz>4cG&@6yT$ME(|ofcTavYWUo)H-uY%Kc#f3KYmRjhV5dnyF=ucsq4bFd0a5g7 zgD{ZsW1*|GqNlOve=FKfWYFQA1m8Q#+ir*aJ{{}Ur*oRd1`cO<5?L&$j+-B^n|AgM z2m>DW?}l%E)Cm-L!^o+5|FB-N7k@NlqPs%{>zP3cpkv@0aSR)m*`&Mwu(M}1 z3)q$-UuJluSv_cxIP~mdAa&I0`}y98f|kt5fP-ZJ8DAa0 zK>ImvpOvtnY`DWRq1Ml!-z8^~a+^S8WSPLr!W zH>t4V53}NVfR5hGdP4>uJpGwW3S~+}^`k(Qpr#ipSmMXV@z7xp=n?V5CAwnrg?H{* z)=Mmo7f{()QTNeLt3aH`?+e^=j;2(#X`Kq`Qt?zAe~uO}5Q?ABeNi)d^+z5%D2DTB z-pXI0K2Uk4hv{1OPL1KzMzUg@>?@ev5uKj7M=N8lMsA;f`m4P#nfOunq6r8QC!VZ? zO&D{-o~KcC1}UVNS38kSELu~;@)uKLsk?b(BIwoedtdb>u$7)41&UhWza&^$kG9hH zk8poPe-#IYV}Ins6IIIgi}7^eo1W-vBTCX_*45C?zJw!R7m!#)pEj zUvuKhxZJF9EVuJywmXUf48)~3q>(o)4emTloD}ZQ+o@@%8ma*p7T+V_0rUELU1mFq zOMWJAoj{rkcdr;a__%U)21**+s~W4z^>H1me@aXzh%-OfEV$4X!#;*(^Fjp_F7u#t z0f!D^s0MpfM-+9;Coci`ZW1$E@jCTi7Vxj?rt`o29sxaILd>WZ7?1i$SJlnb)s6=i zLpz9BIp&f0;@u9)%(Ahf68$&<6?3$(J!p?1fJ@9H7j77FUE_G|F4pjk`<-p!P5)G? ze>d?rNPFwO$(5VKc6QJBqUN|lv+94ViHREA{MbHpuZ(;;+FkOj^%mqGq>q^i1n`1{ z@G4!7+}X%}9(NFws8vXZ83U0zbjfv{(vh|z&97`|w7pX<6e3zyEDWjef9@g=8$vu| z5WfjQXU&QTvi#KgHm@;y&Gr?7ivP3iI!YWMtg}m%P^$Ssl&yd?Vh`6T6 zPs3G4Z-5J0j#A7g1C?)L=O$wcQ>oTL8?sADe@LnJv;m;Gi?-|SOEHdbMH)2Pf6#j) z0~6VZX1%EuQNz}nkYGV?jI<+82zd}{-53U@p#+w*J_h2(S4`01F*BXEGq%wSZ#28K z->Q$lm}!Moir#g$tf@}8yq)(dm~i007)CU(5YW}4>4TaSfuM&wh#TH04ubzVJFuge z@xb{pA`Qoi`ysw5Ob0^CgR(iCe-Tie{z?@#69}*TrrJ04d>c?O|cuc8;}nlATD6$Xg8$NBx*5r4US=(rm+GJn$@m>4?28ew`E6By^ad4~@YI z$R3yKC;n(&ZZw|QOG?D~)fv0ejTin`h)#?6oNtNeg9}Xkg%&uH@d{|QhuSx?l^!G^ zX_AA9Q>QXhj9QKco=bArf158&7F6DzNO$AocqIY|`Vd~)o6J4BuI^-g;@fX?#_79d z<0)^zp1Bi7{P(aJp#Xo2?z*5gLF%p6H~GAg%s)^`v^Av$Tot#i-0gs*l9G znDo4x+1j*0^Zfcw=8;TxKuxkdh-5-rIWVAp>UwgyB-pk6W#n->i z2OCUn6}QR9C!pxD-E?yGZn8#LW4I(!fC?~HKIcxC+RcOf1-mbvPsph}-hi1NctzFF z%cy_b3pL*U@7YsPf5YHH8w47li}wWgde@D@G|D6B`ip~{tBM4u3%Woi7)hy`LtmKP z0`$}74X-;J&snKN`Hub}J;j~2cg(QIs<`vU5ar7bhE~Tm&0Pe^U!8~{&;+5vjK#*X z#^X#th1DD}fIJjlY6HNJ>-ASObq;-ZD^H$7U-fr>lCvs;e;!eMIzEuG-W;$Gag54T z{qA?y%>wX>%0I9&fP2_r^xzsg7irjHevO)Liq+T0vb>z1RDo?zkGji-O}49Ovhh@k!ON1 zC_t1|)TlK_s>23>iXtPYYtFKBu5-di1KnP53ROBne|N}KZ6f^RH13o;IjjpLpj|T- z$BJm&hSQeo`VItkyr>FfJ~4RwXAHW2=Nsenz8f#g17>!CuAYrusdiQmD;wi@3kvSG zf(wXH(2{_3#OxyQBJS9Ks+Z0kwJ;=&zK-5aW8(F;%3~}rrdkC&QxZEhB%Xk@z@2(4 z`o|$)e+&*ZU03E~3S0Dw%3ph@Td)N1>*gdcD5DgaIk(JFsjL;1C-_UV)J@D!jftQr zIW_DXOyf6N1;l>ie2Bc{|;Z1(WMT67R$tp?f@)&xT-&tq8@)%V&+Yd_~KVfQgVu+?YAw(Nf43Pi>uvpCbZ5(*Dcahb1546-|^Zwf11KGgV}#Ub^Q ze;KS)_>H)trV4Z+yQ=pkvB`kc_!(|IQDx$<1L`mAcA!Q%SW%u^)v#)4h$?orO3EM~ zZrur6wUaoczytDvY_?X>?xdc9EMZIcg8HQG%0{qPC(DL_^XlIRSb)$@{{>U)B~w{` z`#PQ^0x3JNZS60igWx9|k-j@{itg_ye-wHkuY!9j!{AnMPwk77PYpnQ)>lzKFBItg z3J#d~#{D|J$X$k{cI$6jpp1-}EHg_PJ`U=_hJ{uLh{k>O3($hY|HiTat-tLEVFRm8 zs}avW^qV`#Bn>0HXs;e@+XNpJ2ynD|d5}h4|L3>Up*6iZl1xd*74bUVlXv@3e~8I_ zpw|fC(?UO3;fhkqJIibybCKGxrs)soR@5ogixp368gf<`?mxbyZ-h5TPUU93K+uNP zu?jBJ5Sps%WKyf$P`VQE+VmJJA|e;_Rx{1eC^y_lFUIed;~?8hu7wZS4tPFdWLlb{ z%Y83KOvMWzcUQ3|pQ}^nrq{^Ze?BoW$Gb@1xd!O1XdcSM1uc=lY(yn>6dpJd*BdPN z4|5kjdS^QIK!z4175{zBo}7W+J7Q?J@;0cR$&oIdR%cWfqqDlBfG$X9#V1(&@2hK5 zF~yddeC3>C8BC+~`!xlYdXqHp&ck}jBS0CtapG@Y56jRLbc_6cNZN+YfBHmaQ$^)L zRq&59s()580mH^Azo*ij(h7G}=ZI=F|9FF_#M29V@6Rl0LF8IAmqgrh(-lL43yVQT zvETlyA%@G08Q~-yep3-Q5L0q|a_tNl;>;pt z7*3^8wj(bv=Wa2E>dt4J6Hp&xuUU8;>UP_n5)h=UYMjX87bPFg!aRn>i?6F$eqPAD zM<)e)L0K+4@tGBD2t>3Vw(IL?zh{XL(3bWA4miy-=h(V^e%RTut48AXCp4U{qh>bUrw)f4KN9m*Q{IzUcbsDBdV% z_#zd=ylS`aui}NEhRpn$4+xVtpMfko4U~UAhKSdJQ)19GKE1us&!#Fas~%)pcH&o_ zM^CID&IUqEgzSvSo5LOCzx{juY${}LA89k!UydI1uks_Ktt^Q`UT&>rMMqa= z$@mkk1NM1rgvx5P$@LljkJ>=hKN|1m^R1t`KMpT1y8em$tDo8(j?Ds^C(OulzB5ru z)VXo*H$$r1H1i*)89WiRo?N>#3+Nz!Z?iDxfPML$e>R8dJ3S|PH`tXBdD%vskcV9E zeQh{R3z`GG%@=p?F>m*xoBLtTHZ*Mq2zBPK4Qtzlp!>2E(~sp}6hFJMT{)Cy?>+=u z)xu&27hKb1;Td?ztK2fMR;F{mjbc&NmXg}mHm$Nha$&@_IE1UbfY}A72gvbZmHBsEj=7z`K zUjW$W%xTQjYgy6pS17Pg8hJwbS<#wIy9kzM=gVu(Zwv~iW7qM1hUdiR+vYY0B=AuydC4G0(s`0vRoD+}p4IESA@qV1vQU{CmXS|S@ ze)~rTKF-1nw02wpvRr1yAMuh|rs^D({=$IDeR_!_ipiuggM!Y8MDdDa z)e;|;{YEzeJ7svG#CO|R#=%=T1lJd;U236+^~O-l{!QQ59U!ruvJgjF9=t`J6SceW zKYT~eQo75L%D~b>d#$QB8`Q9hCBn9dq9bk`UUfWu6Bo9*pb`|Nzs%)Xa>yK&e+SCP zgqdjKE;*_n1MfgrFs*jK3(mTcT=9#6&`KsWvl^}6Kvwgv;SR6XccrMn+Oio%aM(u) zx%jQ$X&4(0Q<$x`@3udr3@+{FUk7|3@5Wd|2b#Z*REidVbW!3w>0)yb)rL)a6vo=_ z6|)#PXu_L>Uyk{`X#Wz$ce|w-f6xDz0qAzar33AiMmvKCB^Py&f~_NzR+&mZa^L*{DH?sEK&&+BM~Lx!D!$R64aVaL zN~aZq{EG!w^D%9f8!cKdv{h(+)t+1mTij2MJ36=C^H2GAnA@~N`Jm|ye-&nS7O|jy z2*|u%=iLK69$Zk#)|T~Z5;Xa-tjcp!%$q^B8-4+vd5}bAY(AqOV46DY#+w#6oyv(> za8)c87vSW*zPvQIP|=zb*->G;*{YNHQ$ty>y@9Z$(Y!)hC%hGyGE4Q3=ETvE?a2J> zb2zrQ$YBB1-acQfqmxK5s`q{>X{AyC;ktIgML7%$br)t3&`-p%OZq=rxN9;C3Qeh}~@CA#gyE=f;> z{S5HuFrX3G7~fB1kjNR(@By%}-|dfjaj zcq^Mj)x*e$rtWhBY0@UPin5+=b9HCGrUfgTO@^RRbpPj6+P9JfGt$+MECvc;G$CYJ z2eehxxqM>*fBr{qf!wyp7+qB=V#EC?8)gL0YAVJ|rl;ejPYS=bpPr?(Yjk43W65{H z$09$PEYs~#oZ3E{pEtiw7O#=CGJLNLMcLt*#CuWM5 zlns&L)nPX`&aiJbK>U#ElWsPB+vgK)&B3R*qtlk^f9W{MOaj!vC)vcjW$ij-Z!B1t z!&#rymZ>mTaSA0*YP1>wg`9q|XQz*(t;Q}Y@hc2AG$3fDrLw}nMhBV(LdDWG^gImM zOVCPRmiMfr^KoF2m34FT<@|UJ8Qn4V3^LGqjp~A;9CoU~e?J7>CpYzvhR&qcgb|#`y7D{kO?QXqW*Mh3tO#3g?=b;DgYdf9X9w9ce7*PqZ>~_DM!D?yc}owcFA*Y8 z9!{AlP?|w-kd6cE)IjpOHdu~au%f2jUBpFvEFqSCc+c<0W8|Fv0cL-At;t^V~60k zA;ZFRwvxbxl{s+b9edigSI8_)>m(P0I0I2}HU1EsGoa>2joet!bSG3Z_|tpy8Q>qq zf0<5u38RC*bB87!X|=Q9%Isr73A4vD8O-w)Di|#ab(7}Au{izXZw&;f+UMxvV7S2H z(=X0Ldu-}NvL72WW1SLnL$VZLqtYPK<2Us;t}^%*KlZM0iEQ%|-Lad!3$i5#sQ{0I z%UPfP@Zq&~qeB1b6a{;g@er8l3@Zg*e`?j+@rM1OZlM^}ylNqt_Ec}<#~4-_N=9G- z2)@#3jLePH+`uH00ZazP?2wWH*iWw_M6gk|VAzVbd{GAzlc@FYO3BE@AzlCGt*fhe z>3?q$>ls>i2v#Zn8H=`&iMmR)t!c}s;WllBP{V#k5-`#|{wH~}><^~Jffy5xf2iY! z%zE(RsbfD3OVCCMz9R*5xe-%p?;dM?IZD>TCc>RExP>4z1Gv*)DE*KLB{ilwRGLvvks6oeiyr$CU5f2<099$*6E+vdsHaC9ht2I@7;0;rp;8}R20=Swf9 zA5N6*6}~@X!ae>X;(5ZUYjrSphy1g>9=Y8h40a&N_u}tW3aWFEO0da9grIUP%%kzI zUm}Avy|S*3Krc5=uOcDK2eM$n2y()+&A3>Hp<1~n(}tF*SEG1!1x%VAf7LRb=|QGC zymTK^PBc*mtjJ}_o(;*81DBd@Oe2V|LN71ltwi-mX#?+}Tq#$&jcY)V4>xA2Byn3C zjW2R!UFXsDw>ot?A&txu%b9%nt<|IGfL&-hZilEd2Evv?l3a$Eg?6S13$M8Jb$AB{ z9J0Sq*&fEf|9(^z0_oYmfA(h5_FZQ~kL~r!u>BLx4j)6aJ%|C5PlJYQ*>&`zD0=zC^&#*ya@DU}oP6(=9}*a$d(+)8gPTl36ff1>0^TfSKxhNF?8 zeHzAwqe$Fhjhmoy>SkU-Hx${!rya@Uep$h9V25Wj(%%a-%~rNy4lu=P&^ySRybHx! z87}Pef}?1qcViFqu&>zJTfrbn29MSLZgb?ge>Y3$mipPtXj zLU@#=u}O>-@^+c@ryiP3Y<7V;Yt_C>sz~Ov?!n}7E`QF-UST<$7}L3Uvzs_x8sVWL zUp3JMqg`jq;QURw`Yy!jY)pd;Qh4WeF6tQj&ci-uBx-o!e>T`LJ_bvNKhORcP&O|H zt;WkWL?Mi|y5B{8vxXf##YwCd;TN?BLxZ@JCsD>Ojn>m?no%GV8bqBrW7i@33qRh^ zqQj1Gc995J$AzgBL~Sf0Y&PkO)?7JSd%IZKb|Ln7{{1K@VkI6i)UUfw;13raAv;Y9f-^*1ONr zUSHCcEt@FgabjHllMNK?+G=6sei1iOm;G&7OnUW_I|R*IaDe7GQ?(PTC|)EQLPIk{ z`1-ySzXP=ZrbTg8)|Vqs;F0WI*WTLOn|OuP$yRz?f0((n_;TbX+04l-dtVAE1=cbT z1)mQy>}Xv@Ii(&$UTxNMc+GDq4z;W*gh9U4Ck|2F4J^81816Gv_!j52_tW)@m`ST$ z=^jASW9i2AvX2J<0(L4pV$H}N&#y+gi83fg>_-JO9f2|&V5Y_|@x}o+8U+I1N|RL) zI;whOf3rD8_#w*uFkpHQPMd&0$-Bb5A32YhL|fXKc=u-P;|84q(oR~rnB5plNjY{X_hm z0SkXHuxe@;J`i!GMBaiU8D|X4f6PqiIJy!U{&6+)vEy@-l~*CWEp0hW`DfIC;|$O~ zx*H)UgB=8J*bDWiCB1X)*)mQ2bPU@Q#%kVAMEkp#aF7?kb6$%_(P?}tun`}sJHh_t z@);;2T<_7Qe;t!1#VOziU17`N_1)QN@6{wV9f!$8d<|cSy}Y&d%?R5&e{%9^!Pc(( z=VJYgq>wY|67;1EKNWre-?zZ-lFHPb(kLx^!v)fXRrrue_%pR-M8w&oIx7+$%S);i zZEI4Pq=dtK6QKx#Vw3o+P&$Qare`tspv&w*4{zU7B5u?tP|@Uk_Z0yOAn{rE;`#(aDDCmB8WY26^oMV+hjRd$s2zck z?YnW+(nnNan_n@f)hRC;@!mZv&<`+@&!BB%h55I$kb!!*`uj=@I5aojeW`B}ACG{{ zd%lJYx%R;@9qGi2e{Ypxy@b^A0ccn4&v`8zJoGSmv%u4jM(xUi5%o6WlieHzy^QSH zgw)QczTIbqIRV=^Z0Ftw&b5DG+>-Cl9YD4e#=dPunA=zWmd)i`Gw{_6#X-E2S;4XrfZ+g;0^Mt93J3y@&RCYFcxJrDecVWTug=@2-o zFwj*FRWR<<1gS6B244ViF}FS$a1)qPA^_*s&U&-{RfL^^-9@rXhy;X@Hy&m^xIF6y zyqNRNcg3(+e_2hNq+RCeMPq$O{H>b$m|*;Zc!U3uT| z_qn#3aVgA1WjJ>CeIPmMHLv0-ud}`F7HeG&ioQMvf7(arCuV1%!^%`<-gdP5$G{I% zYr!tl_hR(C8_5{A&06zBmB}HHFf^cVqWG{0V1aq{)yRE**zNJKh{|cHXYrOU&*34|3}deA+E)Ve8wdHo#x6z_IIkrB+=wY3Z)+@BY&>!!_abz64>de|nl$3Lec5@oXAlhfrB<^8z6=Qf=6F zM9EZl(0xv?r7<w%3P5Ea@*qJxBTYRem)1E>we-}x99qQ5C_>ZT8mF^F&{b<}jiweYk0Vaa0 zOV=rPRYI00-r<+TuPodE^m;6>0x7@j=2U4Q?~ZzT!)~x(o}i6mAD7$l+fcXm@dvds z)pSs@Bbu#L8>SODNRNb2TdO@kWmauYB^xqM4&iCorwz++9#KKvWVA5pYV(HF^htS+gz&?A0l%XQdQjI-D;eW zCSV9}bTo@jhbNjLLB8Kx>2<~GapsJtZN{+s8$x@zoWErIia^5kvGGJ_f5uY3s8Rur zSffU08M3gB-cT!&lav4uFr@0BydDk2uB)hl2wU$IXI~%ZrFf0ydq&tv9z=qD3 zzoiKjl4v!7{VSs2pL9jxO>F zER(6y09JS>t6Y!tavsV&(uGr+m~FD2uC&*xztBcpjiGF(6pXf0x5n?p)oM9W>>_qm zF$AelSB%7zAj(>tXHX3%fc3Kvzr%gw*_D8_wPz4eL#xHuMr8@X7;Zydza}x|O@c>H zmm55zpt|c8p~8ere{r9wd%nasw5A=ctltHG9ByM4mOnv{U0GsFvd;jp6WR@)e9Y;f z(Qn#|&XvxEhLdxPn26tB_v&4BjBs$R4xMHXpSr0R7M)d?t|jQXA1-X?nGc{ZzdXT0 za_^cp)Dq4vkIT)Z49Hnzud%5}=hAtZcm#9H*SOJYz@VR(f5Jlaw^gMBgN}IY$iN0< z$ibu19)m$w_h4=8hn^w{CXaSmlN(NpZNAV+;55oH(H7B!4j>W?x-RjdjeyHuXcjP0 z_3;v$&SoxRRSTmdsNWVpP{gsJf4k=TL+2d(~;|{Pmpd{mT`3fa)U<2RHAdMLL=LA3yRfNe*hYzqmxw!Id{bq>U+0^(5Mp{d^q{+Mk3?Gy@GOXJ5v>i|Ck5o8v~b%dv5-(#{dkXKz~pk-YkvB+8r1uV zdi_nx|NKWR5;c)J?;4gH|n1F>}t6#Hv` zfKAMNHnb4l_&#L_*PiSdOvcIopJ8EJDuq|+JAL~oj}sgKcg>B#6(}G0Vg2e zniE8k^dSKQ_hl@m=5zrwk534ve?Y0JR#qB9HRUm?q$+{(& z4w!(-MZaXuz0EG zt^_9v;OE|0+BG6=Sc_m6vQ;*!NtJV0k6uC}+w#&a&cu4m{V&=ssR;E(GdZQ#2Rc7eTc* zL&N>kgZh9UzC6t>%|!{H{)Axhe^@lZP}%Yy-qY1Os)qO7X;vcAoE{t`PYd5XA@nSo zjvG+?E1*&U7#Hs=zm2vrMscB6J#D|)`4otLkOLnwH#Cvj#h5PINC}eJCP;wI_9Zaf--SZ6RqO@_tpqEy8P`8vVQqjUf0b$M{OOXQ zvG2248_iG0JPd{)PNmb^bpVTPcS>$!r$Y6lP1IjKT5TH^GXNhnB}EDbzb+=+ZER4Q zOdu3=%Sb0>3LAYLs$84C*1sZNLXubDvx&*?B#hFh5zivTbN9$tjt9O_`k|~*or=T| zqGKo=jBi%OE&SY*dhUDqe{ly3lOS{@I%LgHz|MrrYPleSiq%&aKx7%)$Efpo}PSnh3}HB<~}rpX$xLLUhZZXrH(!7Y!Zx;PSi<6CK0S~;K>)*Jn}4l zYfAu@%HtiQf)t`dh$DYl2uXtE(1sCp z1N^PBXm%h4*}=)9W6moIJ}n}-$Y36;^Lv1xWHaA?hrrOAVKoUs?k+z{gV2u=jBWwg z+}-L^j|$oT5JlrL*cz^u?70obLn|@uRVE#&1t~~aKXXwJW77hGiDpE%j0eA`@7QSp z7A82Pst?b@C~xmje;L9#0`8dv=K9|trwcPuJNe|duI=6o`tn+huJ%|_^l z*q(MTW$V?8e^Yh;df!L6WcxDt2PHTNme=HZQq96veH~cl*tVBsY~JCn~$2@^2X z?-`1{7Pn@&_bLCBAH;k zlUUBu`0u|ze`GF3e;p?>;(3VK6-)^4`e#7Ca1h{W;D~n4E8~$Omr<^NdNbk~W%5pW z=d4d9>wbVeow-CD&JK6WqotWyn!z3sSdFnpMqoykOk>Ebo)ghJ8!h_~;lBRsk4cq) zKeG{3D-6=i8!C`i4AHd}ydma2_H%pvGrgEP&d9rVf24IqzIw5-(<2~W)br_Ov=7wiTemEp=@8C^R&Zy+THOhhMO_wtan%@Pg zSph)BC8jnd0TBQgQkWK@3l5jt;0_*7H%lH*{Y+RO0z;)+ZAGhI5IOU&ytFwe!;Ebl ze=V;SI32-`#9pv~CncY(s0H2LrHbyE&@`cggY<%UA7eujwJ{u++2i|1`wc7hc%ms| z#cd3LElw^4thqaV6!`xY5chSBY+j6_)*=>Kyx&teIdF;2f6ADGIMezVP{OgKhr@c@ zmVq84g||KnK#6j8qPbcqg_&@yid4zif9|iZ(fs}?u4=vNp0Nw=!#ym$xP3QJ7p}%HNK&IboCUPNL~4)u^QPSHo^Aj=IprPqu{O+^*9c_Ng#;%6Bp^(0knf0R=F< z_qM)nR0AC`pS>rsNWEnKs0u5k|Hgw)ie-wvZTfICIE!$uKzu0-8%v+#1tQ?C0iTok zo?4zSU64JWQN{viYf^Al={QW7f4DdX3V^J=H{dMu1~HN8vziFWp_WE<9n1rgn4vE^ ztKO9$ywb{2@g6Phf%I>AgJlbH zDx9ENGw{6UXT9X5CNiK7VA90@S%+ezvffTGisrk2N5Ii9dk(_(5a~02e zL$Yx6Ns7$Cjf|l3iw9Mgk< zR8U{_-5v`Juj%H&|HFMOQB(UGj(xRcx3BRCR-@7}WV0aIbEV;ihC8ab|B5s3R*pLl zg+99$QY`6@GtCfEf9|Wq@`hCG5rKG5@=KwF)B$P%FWDQ= z?1r;bfXYWlH9U^*3q>bhg8{zrA&`g<(b2VC&n0=4pU&XA0>ia&G%xfZ7Q??IeT+3o zt8*#wH2JFIzYV&7^i+1f4lb@5Q{G97W!TPWcLX8c zV?R+erJE>LtJPuWgkC@Il47v6$mrF}s1eKi!*9PI&9hYEOp|wOi-es!k~so-wxkSr zHWRoRP(wNzf2zGg*O*&KNDj@t*s#UO5w_tfN-x1w6AkQXH3#*{+c~5mXm}=fBrc{= zQ&im{bPBVoea}V$blwvSl)S8;nT{;Iev02rV90h|4*`WF%!w%zi=2@WeuIxZN@JX3 z*~A>8B}Tzi==kgoayg)E4O)RB_~Km6Qw_gry@)a7e{F)?sd>5!u<4R)C~15b84%b< zTCZjpje?_`Owj~i`=Q=0;y8x#Ux@wqO!d{Ec-iZL<%y>pcY(DX^+&I4za*#zvy{6% zW5c*7FT#&CMP1L(wN=8o`O^m8dwJD$a_{A8wU|ur-RwPyS8DBB|7|n z4Ko|Me~ku?3Ao6W??+v6+oGJ&!dlNF?r-=uPh`bgT}vO5q8CTbd>i+Rd?lcny#hJBb%R3yNO;zwM-A22wcm{KCa9`N*B6u zwBSdz*a$}nWA&(Ql~|~r-s+mEKG_k0oqgB+e?tH>K+M0}kZr#ly0g=}uOYv}S;_01M0>yA^J4YgvdUzx?xVX#7}W<`#r1ds z#0Q{+plo{V#i3K!N@*S0mtenBt|gDQ1|r{0WdKAo8Du8?$AZkSDHSKV%Ghq8bbl3_ z60h2J^MA5%IUb<$kVmPsMOXyh25oA>%rG0~UMgLmB1)EepD zU}>epS-mIh03j+hEIJSFmC9~7SDKHuf`&XcyrGz9@<~y9w|$8}V3B#2yRltru-g2H zb2c&8spFPDe9X$U#fca218|`h_b)E-0W_q3wSOz+VinRUTyh0N*2u%Al24R+B1$l< zDAY=+wzMGqSi_=3fz7tiMRn~$Z|M~vBEb(em6|wC_KAm-U#66~8_)`OnZ zE>+CjfSs(*l|&u#hhb3Sy}%?+PDruZx~}$Yb6e!Ej9Zi$+}L^F{klqncQP7r_--a!Un0m4Ji!-f-9N)P9>A8jsGa8-5K^wRV%uh%huW}T)DVA z3iNEwnOi#o{Vq9$z8SIid5uM^j<$_z7Y$g^lH>3bk4)iM9eB3ZG$+564Fzj`Q89sb z!q6>S6W?50;Hf)Kg{)yvmBg6(X$aPuwiYsOoqx{+&($&& zt&d5o86}4B=<|%95atA2V?Bv;6pOE3E;GLYsl8DwjT{7X5;(&T!8H~>nR=hu`(tnU z@ZB}0ciO=h(_BU>Jl!DBY)=F0*~yH3g}fdGh|wjDLJ+H@D0xCG4<~~?j7i&dWGgbk zaU5m@Ni*jWKJ5!YYB-{!#($ORk2>pb+!~buyqO;C<)(>GtyD>DI8KOirKUXQdI7LO z3>icbr;2?d!N|LmDdd&1#HoRUW>8x(e1TFebmkZX5#&Yj$T3l{iz{+?ZB26Mcq#Fy z7x%>oR+XsG6t#{A`i}Cg--h$8Xpg$~NtA0T(?pDh*-@a2TaD3IDSuTcOb^B3^VS1X z7%8f{muELgROtAn#D zKJQpyT&%{dRcxzFf)n+_eT76yPKKqRIMsc`MqkgPtwIF<1!akx2%GW(7E`dTtNK1h zBta99;4XI>ZEHJRCxut#9@}tNnT;|2s6fgLJAcxrEZRX`w1-_rk5L1TYq0dLYokWijMkQ^QNW_CN0hM~T8WU}E?d8n|zRVGYqu0o8A)Tdc21A(` z5d#!639+ier7}7^IbPBxU-dvagxT_w_F=hqQ>T)KVAvOgm~)%PYY$y?FvVtg0?FXE zbKrAXwXFBTIA>4J5gmyi#9PBPviFZLoPVwzl}e(^U4rx*3HQzc@X`hC?Znq=zp0n{ zse}OZ(XJ3-Qz7ej!I>#^J|M=M%%#AaI=?{+()q#?vQ|}BL^D+b?lYJ=CBGAJ9c9kA zQUTdIv2nKwMs9~+8s%z0P<5CyaOoxM0FP+c`>JLkP#U#T24fOVjoTiwfzOVr`=$cQ2<8xnn=9GF$s9FQ+c=DiQTGj6IAl~c6H64~z{ z+;{2}G-r=!Q0%_Gt=m!x9#4$8cC3^ytALE=!!PJko13ap+jQ%PfR$N4e_;efX#m67 zj`wppeK@(?7`^r%AsyFq4^o>5QhzI_j43)6(1#4=;rDqk0qmm%c}PpqUX6z7nb*Jj zmDXC=o^Q+kC^!elQg2ICHd-s=!h?p_5LHcdqZy72HtGCnSorh0JV_4~nDlocLWop( zfTz%^eJY67w+kXoTMN-89+JVS^uD`6qvs*WFV(*zKRsHPrb%e$oEi)d~lFMmUE)M`;@{2z8UvUXSYM z-V?-YVVYs^?FHNO?2TZ|_K<+) z0KqgllwAuwS56~k>PYJ4`hU<|14Z^oc5mn_Stj=ihesyfaRjL^LaeKE{pf1bJiQ0e zbyO*N8eb-eNsfz@Z7aw6|HK<&+Q-*^ zFx8nJpyKLz`Sf-rNPl$$x8$GHCEb>Gui;fpj!<*!q`&B^(b8ZT`kYCTp^X+VMm1T3RLgGR!sd z)=}Wf-Tgl6N1NeR{sC)O2roT5(q4pmYZ4fwYAyAIo&jy5R~7T^o>5 zL44&8`I^26=rwPM#YCMX^wA*TFV!f}9$QAllg?VAUVowXJFdoAL9i|kiJ>hb4Kk27 z3$is)_*0w_fE9`%uE&khoNf}uk3z0mwepWBmPNfL%zuL#EeuztMN;IR@QfR}r4cl& z?JP^0D~MAGYsGHefU2qVNa8w2&_M$AV4I{d^`c2eS_s6#O$F@!Fhaq z@ju@*-B@#KmR-6D=ab=6&(f?aNNjFPm`a{HOw^ISda-^#0WR-fXQjk7} z?`beZ+Hp0PqaJy9DmBLaSw;CPN5cy45IP_0j(HcFg|$= z%vVrEX(~qP+D_LtXj^{W;?HQLyQ78;FVsn9wOBc!;{Bf=N;bPI#^!{Pn0e}Oji&{A2cTM|!+&P(R&Iuj-p<|4+YU-w1 zDYhDS3}Z>wTwBqbrmJTGtWGn4Kyfo(=zq%dYAhPuT=Dr_l%W`huqMqVUPwCDkm)tL#kz)&7=O^t z(`X{InKC>I75!o5Qm}gtQspf%BHy+0HfPkbL}_u#RUIS@(NqNl-|*&Z)B=YSZs7Rh zD>FS3O~B3&2kkQ2@{m)e9sJe(&UE#z93Fe^NdTaPPLTBFw0RiuW%*bW#=eITekCnW znAbR^sJz`&;C;+02lD{3;Zz>%t$$csa}Pn$xPpZK_$I#(EMy+I|BA8YtT>04Z??Vz z?O0khNw~xS>y7JTgnk`+bN@c|nlk7rD=>b9B1?ljukxq}Zz)i*o?9;RV~pA2h{8>_ zbZr?n^NR(nA!gmqbth4gG|p5$HDa#mXtd_po1b%_=0^VfO9Y;uqGPJE$bYby{cG{6 zfRbA2*^9P9i526$&2hF4yd`>Q*$C3Kd~*S z>vz}KKbED`*^WJm z5J5~_wr9|9B4=EX4_>H4I2%plB_Y5nrGL=<&}8*5PFD=i zatJ_!s^PANXqL-vP<7{_BC5=rpeSsYof-q<0qcmmFkV(-V-6J^Y0yC{J7?G?v(2G) z>+!8g#N=DaSymbhC}5RLOn>g*1pYe5Vxbza2wo2fUUVkpuUi7c6CWv>yx_ z@Aa`*4n_twlUPh%+z7x5k|nn)qCDRDt`oC+otxjVF`Po&;N-o%oUw-Ehj4G+Q%>Ih zW7t3QXmWio8=;?GF-ncd1WHNI7TS0j}yv@ zWVI@~Wo=L}eSb794>Eq|M>i#ByB>q^;Yb?AG+3CvEj+TnqUzzh{*}#F4E`qID*41F zR*?e`8^aa1KoZzt7>`L#$lhdJK;H^MP@ijzu?2>LenK4+BMwULs4gUQOnCIdc?F9r z$KQ{-OhR@pBj-0znwX_Wu9>3^55Ar{@rrx!izON@7k7doO| zZjbUvjTcu^cTLu20o$;SE~qn+L;g4-LIq`V99V!F@ZfBAdixx8RG-nsP+xqx>dOYk zFOr@r)(WT@yMJ#N=oFJ+IjQ&-5on4j{`e73r5Z~iQcc;ozHzx{4z%|NqQgMXs7OU?l z=Csdp7>h(-E>ZmmFA^k_Q~(7QWbEC}3DeHx1ApOgeR6!Og9~A~drKCDf~$uxET`;! z+kE`El8MD}1fbIQ$N7{b-az(~x{!hf%fLZUYsG^Dy=#Nu!yYeZC(F}fr%bOF5s11d zEb!U8u&vS3Wn7SzzlI(E!!%<9%79%E7k-D86o20+V8zY6Vr@zQHBg>{aLHN^xqd7q zoqvLoLFd0EjpSI7Q1o2K5BIv+9?}6wFH6|=eDwOl|04t_@5LjV0nG-1h_wqc1(%Yq z9x1sbDnM$0)OxS_BTEqIQTD_+0#iOR;eJji@(djL8GUuj*MRBGgfw&v#~^W`D@K9F zoKK)kNv9&X1aZN&HU$zR@`EnGP|_<#c7JUsR2?DX0Kk#k`#N_26aI{?zoWi7b}=#g zS^8$8D1)?hr99#E^g4!pnZuB(#*V;~xhNicRo5OMjQSL>Cse@tNDUC;5P21-t}ar8 zws~5tZR@Bl`Rc^Kjr)w=O8-WNihDec_VVqtEn1lY&QqM>WWNdNo0?uTuI%m#(|=s8 z^^dh*awEU5gtM#_V#INRWVz*SBZZAwtpQnPF4;1+xoWuXZLSIDwbTUY2JCt>XTNvS z5OhHkQ$BC=Wa!fXm-FJpS#lUTzKcWiWxA0!^<7SoIjXFE%CuCR8rDeaIjPc{2lc@1 z0Am+thT+CEtwH9kBaEhwiif%7!GCc34X#!LM*k)B1HtI92E>z(dLHrbX!>==;PD89 zLWx`m%!^zrsNCB&e70*y@URD^(&}v@Od&S#VlS+jv|FRWsunQl$^}6#`zx%VV78}- zPSC6E$o^tu58~OLJ7LMVok0IX7U6)>`WK!A+JpU>N%vzN1cC#^TwXfxJ%2RAdy9)! z{hi0M#MSH_qzouPgO_FxEelr3|KoZB!yHRUPqP+-Zr+Z25*hcMiR0oZXD^xd!RQk2 z$K|TT!{7n>-C>6bW)6-RE%EsS02q17S;2ef0E;b-O5Moy#mSo5M)K_qbb} zX7W?*p={f31!X)sK!T88Scp)a3YrL+W&v&9Ca$pT`5bdx@i zc1I85e`>-Fd2jT>Pc{ex!lmmE14Az*BDgx6+%hV1b9c`by_E`& z%wh;ns%A;~28;^mrl?uX#kP!+5%1y6*1_)1tg{i&`}LGTLPLfR$}>^REMU2j5_6sr zj@bpnX5r*V=UOg6$X9aYLbPS@<^gP(t}8OL8t3U2{91%q?y|;nrLdk3UvPh>ocm{$Y_IZ2fabiaptgY4XOMk5OYOxqwxz(+WP;?#aU@II( z<1q!jL`?8TOW){OB=-XDu3XjBCPvBw@ghjT8W_cW6M0T#%_>}OaCbjwg_352keO!Q z2f_k*(3zK>fmr;hT&tjTD)v0tbRQC`^K6}2II*C--f6OMR6j~A$*&t@Kx17~QT&_( z_zU}kaeoaWh`^rOL=RNbD!ZtnYDVvPrhT|>Z{0LMRvy8x$BaC;rafVuJo9~)>w_DH zzAS)N7fxAsc0hYmu{Dc?#vO=9o)LXTUm}5R_T9dEGiScl{6^&9rZD^=D4GBRUVSv6 zVIq-cy+%vVma$_x^f4oGx$>rIB@XOw7T7<5&wu3TW*bF-de>WsrZDx36AI^c@B_U! zd3wSyc4!W|>p041IGqsmwcE0eK?GENiT#zbk@dq8bQV3fNrfo<0~fIZR^!LSH8~X& z8qgiJ%77zP0UaHW1V4XjLgcTAmr-7c4owi_DTnsmEADI@{l+h(e3YFXB40ajrw#(e z%zu8Rl7!Wpn9tCU1Ne1QN*5{yq=qaKFFWCkI)1yn-!melJh8lZ>>t)}b|(gnSdU?g za6pvpC9iUV!tkZ8WwSAn{_l10j0-8iAD`yMAr`Uyg;Byv7*=e?t8|g`Z7vIrBQk$b z;9o^FgQiFbh=6QDPX0*R^ae?YBt)qwCx02YH2AwF;(493oYx|7aq}Mb4)kcX? zNLI-i74xk<4XjX3ts=$eE!wQn3zyXE!pW?GF&ztI@_02sH>9yICRX(Vym)>osE^is zA{WIVsMfmW1?DYwU+LjZ$6R4VJG3xkVEp4NE_QHk2-89qZ9+G2@rZ5p8Tz}eJ z4(o~aeW<@N&r+d;lG|y#aZN@IY3^E-x`b~_StE7hwdPIL9!CpD#Ycbq_4voJR%pM^ zf83?H^&S7Iq9LstxZ;FIDJqX?p0JAX;})&Q5~9vsIt`AR{{gT5JL&5*S2N%RSqyxu z=<80!Ho}7gNhJBd3EUaMa0KuOpMT3tf+JMl=u=I81B{5O20Jk1dWo4oM9Qb~7}UfK z$H>rCFMHYfK3Pvz(fv<%n2>FQZqGGDwfMR&`SN*Nxb#PyJEmR)s$dbdS;FqmNV29t z#eP2X9H$WN?~($b+t{oVd<<17<#f>OjfB1BPx;^f$VA(p}$Q)p`6|ZvRUrr zBPv>>u7Kk}Dr0t20Ng6eS&)3jY@X?IjHjUE1q%H)_?= zVX|BVS=*66Ust2&W5tK%Ud5ap-8on`7(`Jz+3viKW`e4cXs@zDa(@ZpH=f1bc|Vkj zbIRtN<;dHcTVdN#zB_8yZSC4}yHJo≀~k1YgHV0lFYqs6lrjn4CE(oyl91Js^Ca zM%|M9pmn^@b}C`txI$=sVq>@<)kWFnDd05*ySTVAU>0j^EWzBbxG@EARoTHC+m5 zms9gpn6{k8#%5N89^-3oD}?(0YKS8-MFB7=!V=xo0w~KiQ}#raQFx+9brOq7@cHB~ ze!jQdGDj;Pc2#D{LrJ#g4_H?zF({XF5JJC|xR6_Ovq@#aoqySN%H<)@8qu$ZT$KhL z3S-D~32T@mxam~BQ^=`>&@o~eamnB_HeLzdbmo>eAvmF}m@N?#QjF}>8 zk0GB@bJa2_bbl>c8`1QhKO2&$6kW}zz_0NIB04)6UW{7_upi2*d2sQw?6AzKNITa< zo20coCMQ<7$>;I1e3jM4ge=cL9gsc3p%Y!QzURC2kwwj&$q|{~{NQO)F8D^e&6cbh z;^kONh`8P0ZCh`Ck=05AJ$A75)R6!0QRX;js};@_M1Ov?mc86CMVlk$@oQf(M^Qq> zJiSiM>S51MST32pnw|z$w()&2@o1Mrw8us0coPJRh^Iged&?_uyh4o76my7! z8-tLEUIxO1_^N3zW#D6_MUDm78s^egteIm4wv%J;(EO`ED%8k1W|J}`96_12 z`(f0>;(uX~nVHh4{F*X7YSZ#^d*f8*vJ}4BjRymuamw;J0l#8h<{l?x5iqJnowxH32oqOF6zsz(@AT0 z@#e|J9(R>O&Z7v~%=K%ww?e2;iU+eFG4$@X3&pW@WBJ$lVr6pSjx2aIZ0sXYWc9@A zryekDsek+RGlh!)jhe=HO<1mha@$eDUlSu!yy+s_8M(v*Du_<;oydQo%5D%tdt&KU zr+MjuMdCqRnmY=HD6 zOg9uU@wWcDhu!R6d~X1PyM%eWV(CDtu#vvuig-yVY&;^d=@5Iav)7hLjT?H!O&}ER zEgi2Cb0U}j4d#CK>Q#JvSmDMnLXyEc(|;ibF%nEs@c|j()P~{k5qT4jIazs_B9>GhP<(HM) zA12iGSN-TX_&P`ELW~q7U;blAv47@0Cj1{0s;V`->DBcL*S5+cX8oCaEz|}v$lV7| zCCAY{h*jtK->RURypL|(_wF0W9I@}0+QA{-h6?*oZJ&-HombojO*AkWqsI@>4r3s( z?P#Aw?(H4{6JA&_vjSP?CicBZp#Li{3Uv`p|w0|%G z<9hoPPAY=66y532u@N|LS%1S>KI!v6bv&(Vk#!K?U9JyL&9wI&wrtxL$|u;%vO4HV|y(F z^i~ds9KgzRPREaIvvtLoIJy-dwH_xqV^)jy{q(86JI1}liD0L? zb*48gFB@tZLr7y;M}Khgiq+6M7?1@c9gGGwTa%35x!(CY@ww!o|M;9x$G{ZKEstnL z%9OH?!EaF@L(BjRIOqScpFY3ymyqQ<8NGg6M9qtcc4{(pgXgiO^?ye^pxCOV~atlPINEtFj_8YDaX+0j9O{FJWzb~&C zMx-hb**Fx7E9WgODZ=bNYF?G4ci%i6TgQ>(ridHZSJWtx#>4WFEZ73asLUn*+@yJ3 zNVnaQ;yleR9Nhl2hF2x@{vt$S#}>W@|p80fT8Fc8iwi|1_Agkdds_}v#- ziv7Kgak&!vs78ksZhb6dhF&+(g_6~yYnYeYv>GJc_{h_yBN|4KNXCk~wY2O&$a#x*OyzE-FjF`Qiq|8t@IP@acCe;iI>am>SFKmSqS-i8pxAjy+ z=n`Mr$)a#Gvjp1^VG=OkAT`MGL9=9FK8>PB91pueVPlv>}OkD7O8NRxmtvAuxr#&*FfUO{& zZ-0b#++28>y1ci}qd&(v@)P~C(Gi33#;eIq!!~A;-MC+8pl#7Dis#@kiD41}Evf(I zQlQQuONYZDt%hM7T`s9WwM{SPgjnp0mu#5P$`+AjES|+@wd11SWc4_!E*S(EVnVul zomY!umv%`3AMslicn@kp&q8eQMt|gk!+$BtV|wu}%oV-N@;m<2N!-~X12b{ zMK&;Q8`A7Ppf^?yAN{`Q)jYPVX9*R|(61_PN|z3QB_Po69A1?^jYqSA2=a(mvwzLk zj7*ePK+rHgLe&)(!~f0y2iwK9)O)k$>@J5!XOJXr%@hzzG1B=oR`4**SGphn`5+K{ zO(BS^+G`qYB0>mjn>s!_2u9-b?{h-NcBQ!xrKMw-`HF8JGB$(#*rADN{SOt9VId`E z-Ue=thw^WB((N`#wi&Cz_j$P8?|**ON&VlA3vr%LYqLXd`x~2+)0Ed-(ju{ex6WyJ zhsTj}rzfX#URsp1cTy)?cpf4@SNqo9>7{^I5ipUaca=<-=Z}Rlb`i573&j@Jyr-a) zzf7ue?%>FtEH<$3O>fvKE=s+kN)kN>@Oxox0fbpnB-D{?$kQ^4p;^|2JAX!cN^t*x z5@WW1;LX5O62|43F(~8HhK9pYtfbXXqDC5*ei6UyGlYhRspf5G-qfd4b;TCdxJelp z^^Fg;2mFp)Hcn?1donipnzQ9pQ^N_>izDUe|BKH?q7&QItXt2>b~v>LbS><_>4~2M zem%9+;bk3MBq<~BKD}8bB?OGOgJZ>DO<}`N4Gi4nH3va zRH6==adw##>o^@e)_w3?Xhr-+wC0zJ<%}Z^8f{gqXoI3IZ%H|>FuQgpHUn&Iy~kC_ zL!m2V&(!uGrkm|(egM+*(nEno>76iR3r>G400#B(gpRGAIXn=aVSkq=7Ja?J&?I=U ze?*unt}Rjv9c@SJ>eTsqR+;V=TCKu>c9OzhttatCOZ#W-)>uh4HM&aNaK)&l{|&34 z59=?vsro94?fHPja;{$mRTpWNQyKNPxxNhWpI?0Ez~h?i`0-{C1ViNv5l<*O=$A851hahZ3H{G6N-6|vjz=$Y{BpgtTiB(hSYF-zXw1`ux8}gpC z;H#a5?F%d}{nTiWa!blD^9&|#%EHIwrrIBW_}5atBdmEMZf7W@NrB$x{W+?KM(tLK z6)QV9%gwS8wxHfDO^>7Tz_jIWB|xF4wiV)s!({qtm^Lq?e1Afq(K|tZ4%T#2bN9?wwT@AdZv+fU9XDAXWcw$KI^bc=Wo_OdJ zBKFqA=a1R6;W5}%YxgVjhrRoH-y3qpWI~(@KJd1{jg&nAA*#`5(&U)t_}XMM+s=jJ zST4K3!M=iM*?&HVutTIl55aNlZ=EA(G3zB`C0#0bXM>X*Rfnx?2?CJ5-7JDlDqvRx zW6(UGf={gPO8xXmK}mXGZ#PEnfbWNkhwEzs(Gim|MKCO0JO8;#`@2agjKaoy2E~IY z&Cf#osFJ5BWE#rfjYmC#ALoC6XHAKp(-2vm(1CEvRev=fP*IZol~>wFm4^hIY>Ze+ z;2v>=nC%F7_xL9J3;C5IPBWWQ7_WJZymD%p6?=4bQG|DtI*hd;COey@jHD!E0jR#P z__CYl6lE5FbT=yr!(r}y1%2I1JZ&`qgasiU1Haz#hNuDsMkNgkXT_6*M!_CjQ~yv~ z6WwCJxqq8DJ*}wmlZ?Tx0}KuMCmmQcJTTpL;^lTv53ksszNr@bTsNO&?W2|;NG;c` zih=|~S1pXenE}(Y&Q*`mk6`C=Pa+trJE0C!yTGXnKZKbkrhiR+`xD2BR6QrOKt&<< z0Jsp}SfHJRA0AUgarRw9AdvRX-P1VQuxkoc$)i^kA=2@jBWo zwZtaA8@5k4@lbn{V8HXB)YtyQHN<(0L8>#6G3QQ(-79UH6fwP%-PCdNe zKYuG_ME9ouPIBE{NW^WqAK$V^VS+2)D1MGe(-1DZ4XHBfDvV1tq*&eQMfICQ<|&Ot zL&Z8@yG|xGuo5&FNz#%Yq($pz7D-*%{GIU5y1Ea>=Ga-dY3NZKYJnu*b~lOT zXZcE{Z;49-2zHeUpIkIU6$hx$1Zra(MSlv%nIpuEt{Fm(8h+ePz@wFAG>q4NLSq~b z9LB$aFBgA=_+)_DhGuVV;jKF$KpypVyM>>N!<5Kfddd>80UhF!7}M{F?)7CIQkLqu zc@)kZ7!y{DoQY|;9E-O*f883SX53FA{i*RbXrI0#gfK;uL;Xp0Pb=;TKjlN_8Gl4T zAF~7t({^1C`4xnRu0 znLosrCXpsfq#VvHNY7(e3R>Q<4dX`!&J5fZG^{NlJxgso0Dy1 zZ+1nQ8A~c>3MC6e(y1uutU}d0uCOX}IqGudpgH4N6!5lo4`!yv8`t)CWn1!nMqX|u z^A{ubRM-GR*3hH-24?1K%5z1KJX2b2i|JkBTFdxHN5_1UEiQsH_JB%9sDFvz_538k zJ6PdCirJu$_{DufB#B|6?1muUz{Wb@w@9SiRxAsZUMN9?M-_HL`Y!Tk?T^5$Cu;2} zg|2nXM@ejj9+Nbp!$e90o3k3d0PPQO+wZ@5*%p(8<&*2=gryvx<+O1EegJxr-~-3} z2Z>BOxxdMHG8L%SI@Ujo;(v=y+`~~0I`mg}ZXi;jv5LOjB!qF-CpIBo*b(4mv9^X3 zmJI1>qM#7;`YVfH$c=4tUh-Nv9Bqr zj^X23K|FaYyZ2hY1>bG<%3m*>IgwVq+79_#*KwVZXm%o*YFE{zWsCl56H1=VxH(yp zJuY)?PPXm=7gfd?_FVXiT;c~i2$83OkbWppyPrS$$bno<4z< zJlESHCCAfGDlL~99_rWc zfYNxRIE~S^|GQ#xW}A?Z6vmZ1)z8Exg|H2LiL6Bei3U}aZz{+KuY&bG=>Vd63o>+# zS2+w`xT(^V2Bj+V=AQ*_` z!Xkftk1=3f2YI_ES?8~K{g-5WbBeCc=6*&{NGhcMh#b!L?;{rR*qhi*cGC={o~rw6 z|3&Cq9&P22^+S}Nyb(3 zR(ME{S%2poNcAeMAy*(2Gn%d)kP?WAXvxso8l5B@q@?Hnt98xH7IsjKQgrnLtrk%e zD-v0@%BR0ib{ScT7VC&6s04_>`9dbUiGhXUJ6728-{?~A`g!4In=daZzH$PgsW)`< zY{IIKVi;Pb_2>{MW&on$j+$uU!j_-ZFs4%}OMeZ^V-br)?nO*E z10C3v;zfl+TgH{*)5b>nhlWPFdp%0G0ZF>zJ6JVqv52nw(o<$9K9V*nB>oaar^`u> zB!ArHzMFx{q1qid2MoOqDhsspmqo=AOdNk80Q6MNkLG?x@JsA!xA|(;f*iuht}P$s zo6i@s^_L-780anEXNM^HKNi?U(g;4ov1sGpy!)KISW6FyNMGMf$O*9mta$z#*U!$B zu&CWOXb#dtuizs7zdromh-;uCf4DC;Sbs0R?D``iB)o%97P^MN?4x%St^6_RpDj;8 zz{*T;4SF|aHVV9FLrBJ>edGpCd1dr{6Klfvg|CIamKx-S1D>GMv|%@9gBxywsZ7ap zqB96t6T{c9YLA2CSe>X@`5bFYd;U>3S_QV&m42DXTS5pBCwVbEnT@H5F%eq{Fkwi9)J;Foo%F+;2~!?5)$W<*gW z&0Hnkmw8avd6<3cr27hCF1oT!q%{$%WNWeV7e{w6*=yd4hN$@S6VD^)`ZpFj=x5x7 z{bV=3m?xd{F1P`Z()?Bx-z-X5PrvJS>rKY+Mdpdx!hcZw?SEy3!Am93 z?QoejK(Cm-Ce9U4UFyDW=<~a?6Xa37>L)l~>m2 z4~^U~_fxPlg8x!;*L7zOry2|?U4g?O+V+ArUZMxK*T79>)OhWQAt zFfN!#D|a43W{yDW3|oUyet#ft5AZ&gdxZ8GL)}JB1~x+K<5?CrTQB^>YIz?y{6;wy zSCD@>e^-)<=QQ2~W?0zdIy+ekRl%Mp9%<|oluqZ=Mz95>tVn(~r z2V9pKycdMQOK`5#@)1OPgRot=P2BnOpMA9nX;R{Via@>!plkd>u9V_ZFjwf)a?binN zhLNL8f@fOYOTfeEpZBW>s_m9=ei_Hbj^5Soj2rel5m$^B$GuLU z9p5}v>}uL?_f9~6vVZm0NipPBu;)y}D>mu35~8ya=r%VMNgZ1a=OaSkPbxuEX^L3}g5xwmwThc- zVp|0vcs%YNrC(#!HD;8~^r?>?pj^!u1uCMz>KXeZQ@=xdhJQ7}z;~S!t*kvUJdKn{ zKX`!&v?0Sa`~R#O8}9U#g6n4+kE67-$hPc}9_#!XQ()hDnnfeC@u5)g&5dZWDg~)9 zLG+*LNAQ8EBc^U7wKqg)L*L5bf`30wecfms=A$Y-U>V5#r&>oVJ+{yP>iKljmHFdX z3ilX4p}P5$rGF?ns}&tY9-5hIT|w3K4E^FwqJ@nYd|$gPB`uRO(9@90%Z;2)m;O<1 zB8e>#Q-Pma!c+I#bIm$Rv;WylfqIuH4&sDtT5~SJ^t;5Vyv$9d&}V1{;* z?8q_IQCz&?ZSJVPYXcLJOV#nqM!;O1bcYOE?bt6iDhv9OH=*FonfrU+C#Woxkv3#e ziaA53K}7HIS%L}0w-g26IveS)Deiy}Yl+`Ym?@HwE`J5pt&Jg(mkk#CRy^HcPhhqs zd-9>6Tz_Bz6KVtm3dOnH9dH->PVo##4W^&D` zlyYkhG`@4vFhUjX>=DJc8cJ0$UOY#~yx;1pFdR!u?cT4w^O1!=B@c4RpjR_~eVbzC ztRW=8`0TsPtnEen8=Wt z9p`saUCHVht@lfWQyPgz9Q@pJ-Nic=(VXLgp>v4MI~~;~BA^S;8Kw_UKN7Ebu^>>4 zkAJq%fJ?|itkw&p#m5{ByhY@0$+CAayfTsjipHH4tdM`_M7p~(P+gxJYB*!F;Yh@& zQ`W?(_OU_i=cw^CzuLKN#MU(2?mFkJ99M!IN{RH-iiQj^*)-ePj zBa0{OZ4%08%QjV!UR2EDHDN-!kiI*;t$$xR5y4>6FTn6Eo$u$KFmS%bz)c(Udz<^N zOYH%G0SbT)&cSk=cL}O%ffL+m{A1H1M6`5b`^G;0Jtd!^^#DZsQQ@^|X?nOu_ERLg zZ^f3~R5mCwxo_kJ{>g3y`6{^gnxk6%Q$g&r+&!ck--gFBfi*ar@+vSqk)PMiz<;w1 z8@A%iTVq}Xu}4?ir~xztw><90Jel}@?pINK)ZKyj9FK}yvgYa|5IEqWEAR~Lz5<^^ z6J(@}kbaP?Y|pzFZlL;i@I710VWm#-25}M1v>1q6U?nonbASIn;MG(8vOD5Y8eAm}Fmne(%y0Vp^fn%{gyI z{szQ&BH)sm_H?|~Gr|LM*G>5{_{&AhlIg)2l|hTs`eQ~Wqg}y}n76J^TYh+w5Nqhq zYkEsFu{QvM5ZB?JWc?oL4Rs3#(LI| z+@oo2bB0-$dkcA=)X=epIbPhfTOhhz=WKc=yUf z;2cosIwxNS5@BR^wvie_mSG)SP#k+Wz}GBbxt*H@{c7I=@4|6La3;+aj1 zg~~8KI-=MP71A`;EfU5=lb!ds3i=<4+?XZQ6(M3+wYL-pd%CZ5RMS!2tw*Bt_@p?5 z#T5ljS}NQ0L~H1+53!xH-Y0nh7DC9)&dn8^NFxbdA%NrIXn$)jYg70w$yH171I0_n zmxps;zR8Ht+;W@Ssp~}+{r0JW4tMB;U>;!sRGu3Ay0TMz_NM?vK)SzByc6mKxG;4~ zZ5Vp>YFa+A_;$HCKgLmlI6fZGC@Pvu6slrakcz7zFJI-R6TA>Ywps@gc_nBuA_~2+sE&8^|>v zis%ffxzdRq4i^jjWDF&ydyMhPVK?>?{RL?!efj1#B4?a0*_}?S$ThXNQ;0(afTt?r0UB8 zQ^8h??Y0}Dl-01GDp1v}O6u>Wi}|t6<&P&2iOuvsr(ir^$q5K_zl+;=uLB$xujec4 z;3%V9@iScXq)LZj`6-I>2fVHzhp^+^4ShJ_zJ<+L}@p#t1S|En-JGRMA{9lI5N2eXbytUbQAnbhMH` zmy3snvT8DM-IvzVUk5Qxr@h$nsI6u_)%cVYWUjopz)pYUX{z%cv78bZ=zEXuHPuMR z2)X^dP#bh~iA>C;6Ewa2AHRPY=KZn8cF%2k$$1&- zp@bBo?*1hD5F>}J{f_q&CKwsja}!&uJjBX)5}2z$aZYn%7Gh`DSYmm5{mOo@^@I0o>?oAbnH?8s;}^y3ENN80ieAVT2sXUgp=SEenVOl0E=B9cYQv8E)A&7t zhqJB2dD`>>+dX+=qMzFwFz*aJo2MRIY0c2JRDXYSE1$4vq@kn)gvzhQDcFn`wVE~a z{O-2sF0ZsIcC;hq^d{J67L1VlBr(K^r(drF7Z%gRpWMzp(si#`8o2n&@VJSy7tC@P zul#)yfDUoWw*yCMTQG+6gKM_YUz#3H=ko4*^YwA@96;pcJf!8c6p7B5PBZiob{c3zHO7q3S8u{qYndi{cy&ILeWkCRI4pUlQP39&KQEfb$FQRk*bT?sVcM zAV;PeUHUME*Zljq?wzL^M1d<$FN67@T?pSRk1Ppk;j>ZooyA*plt5q^nTwFa{b#Ke zqTTo=k$zPnKRJ}W+HTbN1xv{@b4g@U*~ou3Hrf2$NrEpROod*a6|2;tC@bo&afr^B zv0md>Rl4fP0Fv6fud;D+_J*n@W)3cPhL+77LNW;Jsd$*YmY5`Q-R~pa@nFAk@0X z_Yf(~Q^yl2eL0%6uK5!tBQVe3X^HE6K|%#a-E-SDr63vyW2StSn`+@;WuSk7NdC!H zOXyr=#s;<0qFcZfg;BBy>HLDw4aW@5h9SEiXF*I9+xh1h&&x0%@7lBT@_O3ea>y}p z;Efsi1KP8EdiU8AG+VppI%b*7Tg<&29=wUj-`A4g`$v}JtqNnZzE+h@fmtk*)-noo zA5F~niqANr0eoF#u)C-nwn&0V3V@2yV(}E!2`;atQO@WQh$rlFBV8icDrVRoKu&%`ZTQgH4sA>k%I+ zvExZ;Xl*cD)w^{0N{1!nGNgtPz^C}mDJ5@LOV08u)h~8?;d*?jFaUou0-@q;C+Jy( zXE@F6sFhkxpPdYk!FRXQD+f~{n2)Sh)Iua(G$)xs*Gs9epnK$s0#mTwGf4!{*J~m| zr%FW{?*1uf;ugci4|s~AP_SF{%O2u%k~35_*Q>ty5yWRn*4py~Hqs`19|XmPdgRMp ze`EPVB=<~v>emo*%`|^bmni3}jnWQk8pI9Ly6I+PTZy~J2|vLF>nFof`+;r_AkO8v zNw_4sBL1xAQKD~YZUHoxo#pU`33(7+WI%di$E==4&)246&4f20Kx}n?;YQ!%l}md* zF31VtF61|JYl+iPXOUsORlUJWpb8eob40}ILcvy}0K!X@DiD99(;OLF5={0YC`@u@ zjg@R#oe%m->$pJ2CaI!!5FqTrTHtv4Rbs3+AH(59;5ceesu9AZA^0aNekJztk>$9n z*)^d?FOs4{;zX*1s}}orzY4lvYtEMo0L~I3&;B!f(w6qD=KwtUUO|nT0kMc|^Gz7a z2yNm*R`3JS9SVQ-LFRUi5Myk1ZyKF_1d=$O7 z3n!Fz&@2dbDpoe!`VxB{9dPbTBm+{z#*R4Ha$H#x<*|sOZvg zZBuiJm|K4}e4pF*v%6K5m7fR>GK1(9e2fA-{FF#m)=`ri#mc=fJ#C{+&pq`1EnoF* z%_9%6vO#~l{PB%wPE#LOB21v%`a&NoJj|US=<4U$RV=u4$EP7bT)C_(*KFhzm4)~N zaG64Msk(#Wd(&>+utmp!jCdKJ-;zncKE-q@ooYA%OQOAe(qG&AUDx-hh{pRYuFWnwmZ8o0jd+FI!aBhGr0F$ z9`qPmCOvwA>P|@$>ULbK%gC% zB8-2R6vr9fEpp#(gWB0RX|cKP5>DRJOU^>c)>P=vfs3Av`8MlPkOs|!YRv%xr-IB2 ze2xA;?BGpdV!&l14}_pVAMxP(WdgOZUzi%MnM(DtGJ(!Q7`{B)7aDik2k*@s(VEH} zUWtya!^@R@tCUBCS{jfcb=up6r%bt)^_G8_#8*_-edkV;{drRgJ#4>5PyZUWyJb5nJ^t=JxNWeY0-&(W1dRADq_*n-ExFg3 zld`gonA`o~bYU?UxU1a#=Fofc2Pkt0p^Fo6+>iomL(goPNeact!TZ!N{LgN|m-&CL z!%R@9(FE{p+*c8$g!err4sC*Ugy>uCHbJ;u189G=0XE_x2MY$EjRS`N+UypHtgzo_eetr9uOac4-BAUC3Ju_k{hE+&Lb zPGT36qaCk;3;2nb^I5P^V7qCZ&kT@hg+>$)$9z_=Aqwqc6c3FPcTga-RA&MQ(|f*^ zPUZdDAzQd*q&`+BC55}{4_GIo*fPJ`sAXKwfF6N&dyYuPGsw|kD?Dd)+Q18>^Mi`L znHi`~0ydmK=q~HqfDWjMN7H|J-J8F&mXMQcpa{P>dXadCny3M)u<_ZdS@}`4WNq9<0fj!6VS`()0%uatO04r{QeV`!r zEr7Q72)Js$&;*S!O}Qn{nMRll7uW;#K#7Vg{7KNk5t=uQRm=bV024+>dZNjeeAuhx zvrgAC-Gzy&agEop;YnU!2M9f$i#awQAN|1L6cEPw69+2rqtdY0A;!8~dW4}m#JFz% zB=4~YIZaz<)17mEt4LBedpKwxNRS(+)SIEEmdSY}r=-L?Km@spFC3j`0Dwf;@$axzD zQGl!LP=e-`nlS26oN6$oA!WI0FqRS|87|E*R|GlXCBxx*A1;410?OE+0CH5$b|gy< zLt>bg3~x-y(yoRM!wUef{N|I4r+t^oD$(U5)sx^M$pZAR4XFdGb@CFQt}cM!4#uhz z^;q9SXp9F5j5W4k*o6eS`M38i;X4>Q!|ju%qfB(OS3HlXp;D*|jz=`f()V$Da!D{i zY_6T^sJE2%u+V?`GWDiqCFIYbAGIegZlN7$Juse)cVY%&NM1royb6LhORr3jz<4M% zWE0O9z!V3o_D04L7qiBdrr*Dov7i$2l)P@o-C$uf?|6KG|**!M%t)@S_k+u{D3K8%*Qs2j8rBkO&_n8FN3t z986*#5OqcLrDHP6)qMFmUlf|A0{%d6Cw-j)KGHszs~QP%r1MWUZAYqbtV?NE=K5h< z*E5!d1pRl@m_raPZOSr9k!`71pIdw1Zpz!Z^eQrd+U+*urcWz z28P}Hi+dDsP}kd)h*!!|AGA@5VNv_X3KhjcWln#Si|Q#&&n$LPp~kV2L@2ptmA(yxVgQMzk}gMOornDBR0m=93Crb+6| zhYIwF0W(K+UJrhh#dwPnOKaOlb!+`ryQm)`KgNAs&Y*V8Z^BHbhD?C_QEW%9 z(0+gC4I(KR+oW?bW$S~_R+4ATu_6&b zuNRZv^nDfHqNOGT2t7-k{yN^qn9yGVS&H-xWWbe(VzR8NBcrov4_9Fz^?OizgmR*t zNl36!zk`#rv!S0Ng%U0Hm=eomO;*X*acO@+1D|b%4u8GWtf&(>i1F;!4;A3@&*{{r zJ%yv$Y5?sxfHe4dh{z3_)AIZC;q8~wXNK4I^RC!=jkF%{%aq>URo{#-(#9;RmT>1< zw2)f5Q=ClQ32uM>R8o|hE-n%!3-yIXakQ-TV1;Sw_iEoMU`pcUL2?s~G%j8U)p~!U z0}sg5A33?`MqMIqpg6uhp>^MDD?eVF1VUx{U$2-ZV79$MmK(O;PC>t0?dZt^13wP) zF9T-}N@t!=ey1$igg{B597OvDA*X4W&YXIs$=RTZSb=lHT(f6Q-8xkfuvxY%j=OIH z-{0-IQW66}w}Vjjg?zI<48oJX&kldGa4Rsr+TPSeT}y;0dT}jI{Os6>AeV4RhKqcw z7Tdzl+uh$7sjNP%oTJEEJ_AfM_eSWH$zCWV-B(R-=do~mM$SC(PE3B3e2?{6>+PBN zQDit)o7>juYmJW{zyhTm$Ff-2hnEuKgRp|THROHgrUQ=!z3N;(;z?RstmAiw{~clAhR+g!TCnahUi+H(1n7P%8M%%#rW$t`5n}rn>a_tp7hwjulZ4K z^kKT=W}mSAwG?XZzWDH3*r0zS`&%W;3FXxqCZP8N|0q#gugNmcft`s0Q5ykfh@IQ31A@HV z{^LUS$Id-ACNgVUE??jsrlYRx@K@an7Du!dM9*qj*%)e@hJ4p*_M2QA3>~Zu&Z=vDlMkn%PK|%~4+qeD02ZfkUE^p; z*-wR0Le-Dw4M$s5)=3JpS7r1TmTrN4Hmy>B0&4G;czA6>z{AY~e`4hNQ^KNp}AvElft~yv5e?M*@vFG)ZTv}>C+)ha0}Fiw+?I=Pi>@&P@M284q^UPzr1lVj(sM=lZ=ZYwu3P0 zxpOF?`VC)EbLyLz2XG}LHNzh|;rISdR_m}5kT5A7r8hfb|UK-a}{@I4J%Xsuz~G1-wHnQox2Q(lvi7_Q_+^IA2r{bmDdTBbFw` zghc&JdMUOs4)0WU3eNp@A`8>-Y(wxm`oc|A0Uj#zl?W~-eZK2{4SX*HKcp(iUGUu@ z{pkH`ECk0&-T$+bl-~jyNV03R=4`7dtnM^0~ZbaBdWZUJ*cNzSZ^WZg97l1l? zU}9Vr^ogK{iVfgm^XcDj1BZ1UMW1n~CX?&O0rh_i;?YLN@WgVzoTX&E&3MDem%5kD z+|YHLzjVWPd}S(coKtMYR}w6(Qer0PNWVU!K)b~m6=Dm(={dJcgB=* z%mYyrSB_AS#zxUTW^TD@uDB@wZ6e2>(N6+Ip9b+!lWu zVv9s7pvcqb+zJ$;p=QpVnSa?B1&e4`UWrFQPEz^Ig zR%cMqA7Hfj&SI4aW@g6_u&P){h;6%B*3w`cP%oZbWs zF%j(POA#RE?xV zPvf{Lk#EgNqwHE%@e-M8kt(~$*F!*5KwGxjZw+Eui0oOqtUlG%GYR{bl@J_AYjZbZ zVXZM{Z$+K$Ak#c(X3-T1%`tynR$Y~nJ7&&eZy?o)b9le1`H;7w_HSPT@+i*&%==YN z8z9=jI`!cMzhdXQEWE{Y&<;D^#U$pGlmsydy38N5O_stL;E(S-ftjMNPiJssuJi^h7 zbYg4t-1ck@ju3~73a)XG}=Z3!1Oy z^ybs{Ge2o?`?dkZ6M-7q^Y#o4?9>i!$j~MgNX!q(nMFciT>6%+UwduG(xZ?}4kP z5wqisrosb0JR^UFlA{eaxvwaZ*bq|Xqb)s?WVX}70xsaU(pe)e7t^@{X;Fq@R#;Io zV$?EGgtX3Q@YP<$cTqU>+RzZ#+KLal^b@5TY1v{Y1z6uq?j8VYW*B*Qv3kx<9}fUL zQ83v%NKigoqM>GnvgLU$81K+OS96qWAr#;Z zj{eR*02+qV!&KnXW9uJ0XBP0LQv(*ENahIz&&L+#H)E?PGI0w$0Dts zF%7Ea@%w+6FaVCsF?(#2>#~Nh^FEy5H0P6Bk3jn3x(@uJCvGW`VhjQaQ!Ui53MH(m z-}s@T#KWdsY3eZWrm>Uv_vtqT*f54*i8<_WKu+LC8^^7Vvc6km_HBZWNdpYw$pN8i zyqi(iW>1PVYh6|{10|@maMYVc8ri_XqlPoo7N>szFqH8QFq^98#zmNF)@3lwQ&Pu> z#n&FjW?;0l0dK zybefOI`{|AT?f8)3ewRsD}lvz)=mK;S}R@6K$sgK*D)yj!3O}<4yG~V?X^E7QM$1Y zCvAW9aqj4mT;65x8(%#E)!7H~A%+@F9m!EKcMB;K{HfpUSPS1iMrM0;i1B?!isbXg za^S9wfsRiaSMc(X)P`gH$}V|y;k}dIlUKz$ZbKyuY2!uTh(Y>;)H1WF`6S4{^4;pw>VVpdPKFWX2 z-@m$V7!MmZsB{71$D(uoU0$uzqI-~w;|8x@&A(oz&mG$q)>42K?gom6;sj5<@F88y z(dB%zFXQ9_w^BarLjH%Duir&+K4E2^=4fZYqjI;@w;6NJ4TuDM)ylA9k(MFY5ntZL zPXSs_U}dw}+(KF1yMtwlK$C=v8J2(Q`ekIkv#9`91^PjKqj@F9B*y|Nv3LlpoHlJq zQhwlRqnHxxP|S)E8mZ_*cFcu58g}T*pOeuZbppXl?u_;Dp3wLGV%_&32$P85xz!q2_m0?W4 zS-mjO73sZDn%CM}^9Q{eSe$?S97Gs=-1F!Js`U6(+FmJ7pq7-e6Sz#t%ptLb-8M*> zMq(K7&LP3YMvEJkdKU7%T-h8t$rAs{k1S!}+x6aNK{x?ozMwa1aoqbVigZGNL1L&^ zRA*-{Ddv%DU~@~Z*R@Uj#x}^E)$~9f6sU(BFX=po$4wOh6nm}RdnA7b{{{;Hv>dqR zoKH3CC$W9aKW{d7OjKu`r7zT0ayR#`+IZ%Ev`ny#f_ynJRj?=!yZ2X>@iemJ+*Gqz zs$lzxn;-J|&Z;SLL+9^z%Soe~bQ42^a}_B`kYlHNv;&zoW@)|b4Fhvm;hE6Jpr1~s zt~p|HO#wM;*O?Re_ziz5h{mgqCQqm(cVq@tF*Jp+!-zTsvQ{$~5X%E`B|qJP3QEWx z)Ve@gEy4m-GDu%` zB4*kB*5g9uR#r?FsvapVPq&OQ!t6gvWXhyA&JZ~x{-lDM;?}8>XfOP$#6C)QfE^IE zkT_k>5PmLM>mr4I&mhDCFbFM?=>x;eup$I_R#(LI9#cAs^7U6}pE~58Q@`u-)PHEX zr{ULKDI$OKiJ}YoHZ_?v|BojLGY+3_TIjx?0?_>}yAw$4E=HF+@kXe>L7Sr`WO2CV zEPMk9@BkIy7$M2?!#!EIH^3NBo>bsD{1bg<#5v5(C&E1lVEGjcuZyCUT- zS<)Y-@Vw6TYoyPUG zgwTIPtzB&IKIhbUQO(?+-QU=j2Dsb}K<;ny>2BN~ zj~cq#5tP`!J%ZsWC5m_prBsp3edXM=5HFL^uUbxjbf3;aotkX$zRlu_tFDl?*NQ;mA~i<3r`UX09RoTL-E{+5=y6G#w+D8}A#SOM_6_X4f<*aHcJgMp*-PN0lw zMvwMj-0|gD$fM_O;Jy*1Fmu#KVhv2;1%9dt+ulTY0TP9`cI{`Z)ZQ%J)FG+>YaxG! zXdBXqfmsvXld7>l0s5&LD+S*Xc23G#PcEuW-U0;P3BYq=zx;x>`sJ(nDEmLHFEjsdgTMSI%|X_@j!yq$RkX+X9eI}}s%QLX&DEO*w9d3Ut{{Isv zO>s;uD^+q>K#Qbj4e!W=XZyGLaxc2+qFY0PcTV^1y%V7$-IrtPh7HuwXTKt>>@d6O zev`_9<*xMocVeYihBIHK8<}k^^n3yL(38 z;$52NY_@AqxC81VKCpm|Egy&mJ4y$WIU*T^joa(o@z~pgKLMoo;<&0pFn|Nj#|qOP z2P>+BZQkXcw9#yld^^gT8(HSu*=AbehIiD_{fN2a`h@p5_24^%^c;WvPJ(@fJ{kMH zRcE%B-Fe&vJ!xShdt5B*m=z6u4f5r&zDDEPd|Ngn2+kRp33nZev!ZCWi&Mi6pB@6%jU1B&Z~^c()1M~H@@KC)rz zcdg+B5Xwf93tqzk$pf>;9Chfx zk3^OcO@~)nMOTN9b(jcFb`o z%o5~y8pIrN*+VKqu zLKy=eO`X{OZEmbx92d!4?);SR5_H$Iz?-Dbcqf0er`*heJhlX69mu!*li{O$QJE{{ z)=%ig3~=_IuNO*MUF^rcurx2z@wL8NZYAA^md4!bz~ zZwoSmG$ab}mK!h~b^{KW{0_rJX-717#IRhp%MkjG(N5g$*68rHbOU1Bh4w-T2Qcso znh#KUe39|s12B^2N0t?0!JzO1Cfnra%87sY5*uID;HV?5AVk5`c+-DYbwe4{p%t%3 z16KDyTiMrxChZUFYd2xdv@V^w5W>A0vJb{mf%0fkKgh}rGW4~-&6V|s=PHz_Ykt!` zzll2M$G0FbU|<(HlJ)5FPuIXdLds?u;LppIDaXHSd_AU%QTz~hmGPfxAe6D0CMJJ# zri2mv<^+bp^(ff25e0CJkRmxGF}0BZj-Y2?fwt&~C`OZE1fOOqsfgR*uMbgNiqFL> zeL5Wju?PFpsT=6Kq;4W{T+JHI?hmT-r_dM+0uY1-!$VHKD`O=&9Eo1w5thGf5OZ}l zvEmg1;<$U7Aq0tNf9k1q0X6r-!@_?wRA(gf70$8Z4&R1y49Te5de6*Blob)jBCD^u z$U!_IraK|UaVRtFaA{x65@lce4}F{*cj^MrcL86QA6o4sToTuWYUgY>@uOY?jv0%= z0Y~l=V%r@Ud5B-nA7h%b40ppNe|31)#xUz8uz~{9z~QJ@Ba!}2CR}6!SdV|#t-)5a z0dei5yk(t5^AyF37N1~rbuCc|L=I3Fou#g`8b#15yW!7k54;dvQhj=52#&Z?iT0>Y zjD=R_X!SD%`&q`Tw+zo3RYUs}qhQ=u3amMt8G0}Xz`zDO_6;p2^b<_M!zV#9j)r>1 zmy?n!W5TCk1-qKFpd{9STd;o+Kg4$F$W%MUkh6$rd#X@3W#crRIYexF*JUCHqojEg zagWec){+h)YYxR7&#eya%NjW7ivamwMgZg#-{LL!Y)Y4azu((P-X8bK9bp%$TWNmf z;;}N}3m>ofSA+I~u3j#i3F^T}aM3L#*6XJ0#=!47}!2cW*PdJ=)t zi;Y&wFzw)OH4htNP`LJpChf|c3sl^H(%gm!@Q}PS)JUALHFROlyloyTIhN;elj1_8 zGOp-x#?PAcFN@YdHPo(~n=c*x%*`c(JsWS*|3GBWp;m!ll)$}L$Lyl3a6BS=%K{j?t7%@1!IN1}Cc9bbpfzqv8T$9F( znr0WA7@F&3UoL<99tGTsVL8BPoN>x^Z^bhUu$;|)3YSQE41LxF{;~Y=@?tu%MS6o& zrft8R4*a@R;)N^|IOCvpqIa^#SF;Dm3|175!U}_bmf~#I&`>rb`0j{JO8trO5#%9|Wh;0g=7PoY4w%*M@&;y%u8sKmA0>Z9ZE-XRZ@NY5+p<*Rb-Cap(?Ft0aI(t1Yl^DaMBxxbC zf>f_)jz*T|(3NRgb7!rq1!%NHb`x=DPyrt}?>_@TyQo!-BY#+$9@cW-h;UIEE*~d0 za+&vSVFZ&CnvdfKqjmQ9M*@z$VNj#aU-rvXFU5alCAH&vn)eib3rKnd2zBLMxrEP8 zh*6gM4W>IV&b7lWvPn5?>vLbzoDlkL01O;)rikyc7oCIp9F)tFu2r&Twil1XS7r|Q z9=IVBnN5~waW-m@=}G`~PwBo>jO6wtrN;O?H7-SHDMkHbE8)1>q^> z!r6aN{X?t#Z0)Ri{p+F8$$nn+96PRtVp8bJ=^Ev(;}4b~AEkSZuVDd#^zdd}^p?J|2HPeO^d>P3A`52)Y)>FX!SBC)Ysnls|v|l%@d|`nzS=M7k;m%rL0szAdJ|XZgdg zz$7<&b`$u;2iP~uI+Ey6s7rF!7_9+aP<`1~c}!u)X$&G_j;f9^yM}UuR{=X7PDdDl z_uY-eV@HnkkoU+sH_guxT{ZXuULJ)o(Tj>i(|KEXT4e+k zbD)LP%bR^;tGnCa^$D8-fXv;frlBR14+Kxeo>(f$okZJcN*wj_2l>psX=B!{cHXcE zQm6krwNZ%;UXeg&>5r*E-abGkZexEys^<|wz;(LRk<X@7Lc5rtwp#BV)qf4`>KcE@DL$Tgk4Coh9lhn z8h2l2?%=jRp+0Ch*QNpS7?i<&$jg}vD?zET2uJ`w3~*e;&}IF`FRAXc`VW7h)Kyo! zrWEMfv){eYIN&&lCjcgbK=vU*QY$R zbK)VvEZ2sDo*nuU`a;Eeq8I4BA$+Cybdcg21ed0kL+i(3u`DxldWUcSnAy*-$l_>Xls znxEDAzr@&|6~G8e^QC{XV!#Tr*s)6`0Yz-xP@tT3R%)R5TKb)w_e&-E8Oc0zbHAnu2Vt}g$*Zqzc#RXM##;f9hA})=R4G8c$gTqQ>%j*Jr=fMnvf;>skjlawyALrixtynb?5GrtycZ3 zJ!);{84{*`*$01Omqg-XT<1L(i&hQBfsBC%eJ?}E#j5aDT%Q|unWLyo;JC8fQBlV{ z%af&48U~~SLQ`JG7~N0;1*z|N0CUZ<-w^?fdK| zeaYOsDq$U-VXc;6kaluUXr)v1ivDnA{1CXWBWVNNO);yHPyQLcbI?CAr{}&W`pdWO zqF<+zHO!mN%j>o~#cgU~G1m~vF^Y6!2{(Ud#o?_B8S+7Ft@Cb_BcSHr)Mwf9z0%cN zmbg*O@_RA&rWYxL$;7P&tYEG@c5p|@h47Ou*P0uzOg3y~X)QPc$B^f?R_K(j z_?tuTNC|Km*pW^xU{7IE4k`F;V2$fPt{`alwQ7`*He{-F_wMxhUFWSn@|kH$v)O;D zu3e6aM?{`;Clt^1M9S&4+S-;8=EE~B*5lYdZF)GwIM*=s$&rTy>yaDuha9q&rZ$J7 zHyqX%8Uunn_+CFnFHyFM0Wu;+OCIBvOK^q_x$xdC7bDmc3<6Y;Z{Rmj~ z*Jeb;4(iCR7~m6G{-Uf}KAcGSK)!64n@}c6+2A@R7l3l2Ki?2QzL>wjl!t#2{Mc8L zGl)Q^pLKhsoAsi2%{BUUaRG%v;r6Vx=ET^S}1ja)~tZSiHOwnqt-r7 ztW^2c;wSx|%kk9!0zxO2b%wRPo= zb;|Pk=thaJGwF;d%4c*;SA~D1qIRGKAd%)2dXPp17S+uI`aD42j+*A&&A2u32GmA4 z{hTYN`4qn5<2Tj@S{Fs-;IBihfX9QZac$$8|6GOTiybFjhGUyz#wf-ZbJZRLLr;x& zf7g=Xq9*b{^Ds&0${MnS(r5y>c#^*6j5UqHQ_f$FDc3wauY8l31tWhhryV`nw1a}t zZEjX1b|m%eE`!1jGLm{_1u-KJ36dyl#Q6ZR+w=rXiv5p44Rins{0!xfDk!+1=q(C# zSFPmFfUVhHVzJ|;(OdvRh&L-5t@>?uiN?JuLB7_xn0Vf8c~?GAosiFFcpQ5-{?T)O z*w!bzp0V2wQYo?QSUrCLn^=#=m>A5uhowyzAj|$CLTx5Xx;p2e>w*>R^8F-f0`;uJ z(5x24fis_K9Ahy2`#{$H^pzma38|RTn~1Cggyl>fwEeKWLT;|j>$;Gb_q1?Y#F38I z^K5b+s$9R^QDa}UvxGgultC~0MH$pQYw4p_^pkI2 zjPL9p)tkAFlZxr#fPH@UJ%Y=*#t)InUq1y%@HUaYojiXDI#+&MZl_}%a0batGGdgA zjUKQm;uL=A8QtHuI4I!@$0i9Za4uY)w%tB7_&DiYY6P&Z5ZW{ zY^Q`b#RTZaEyjmRv_ltt>H^IjvN^3^R>6P9E8$}pgLrT~Po!2|kX}JQKhqlDAi`~Q zxFryB0hfPuH4-y>y?u+)PIz3p5)^CtK=NK|zv1pz5@wi`vyYiXj!)srcTrEvD;pSgf&XVVSA-+9g{A{K`V`9e)oPp* zNFcp^8_rk?|7iWsG$S4@j&|!cEQRsMkENRkjK+WWdnL&JnnMmu+HPf|Vu?hEr0!0+ zH_S&+Z1g#I6qiQFqjmMlVLMxbO2#NRMF1Fi=eAaIR|1M(cIs`zvGfYFxQfqqhGWj! zq2+Mu^5JH48u5-mP=Gk+LEif)w&krw3Vl{Ototfl;eO>AQ9K%vXSuOm1fF1yv<;M+ z7!QAB=1jln?IiB;ImBrGcOz=fnc99RgCJc!pi8Ht>aJk`x$3Xe$wNL0CzidDa|>$l z3bFsRn<`NEFrE>gc>Q5<^QUpn;=6nb`og+y0g!sP|aPQ3F$lcACKj zqw2?vP|~%(;@0ajK{lVpew0sb8v}(Y&n!t6??s}ZjGDn6bA_D5!;1VFui_23J%)dK zpI}hVuyJkVP+GT{G)sp}0+PjIiditHh15ZMUF``-$N?Qqdbnz=y9~fN=0g!yZQTeS zub)D5T+y^R;t?_g0nQEh4qqJY8$w+h9-^B!WgFX$Nz#v7NCBK;XI zW^#`N5dEXEtM@^L6-2<;1M=VRLItjnzieTTOZIZdY4(5+h$iKEJn5{uD&4te!wxtV z;N;_dl5`q)HNkMSnl9I9(% zh`l_@ax|$^*ZPtV2ZqR9ws4&7(LZO^fJ2x%CXz3WTj?-zVpf&omD)qr9P1a&-h~tf zS&4k1G^6yB4d`+KHY%=?)#m4vJV${fhjJoGVDRaYmji;d1o;A@k^MK!Sujd}z*tfK zhOyP2F=&8jo&ATvC#P*ob_&?+bfKx=btU#L>e_;;EFc^jaOf+0dUbpc4+pSn3<8^U zA!pRiR<106SMgRQ8pvmuYY%4w&pJ-)OVmYbX5Ply-#wGhwPouJKT&zsF!aX)vfXr< zcIEMwibhZE>*j`)vqCbW3{cB|O7Y&qBc@Tj;O$U&ZoRH1>WAsdDD`?F2K-7)^Rn_3s@6Q>gnGgR#D#XE zcO7qr_N?N%BHW$?6Hf@n%Ly8Rk9P-B5&lI`X(5xo#-CGYMlYBXJ#9CCZ110O$J+r4 zT!S9C2Eqz4G53NR6{OjX%$5i$y2j@Y)9X-#e^Z8Yf;x+yxi88tcq}&s3apQ@#HX{i zU+(jrZb^jsU2{nx4Pi}gq%5q-oUWKPisga0CW)`vebZukKBX?qSUMdTM?0XTe1B@E zrQ>4CXoqM+e#vJV=8z|US4Tz-UeFP_u)7j7iCQ=)B}!Ga^N%pNnb$g_SZk4n->>5v zUn)Me9JsXo7yo$GdS{W|V=O4@|Cphy08i@FGeeZbRkx}kg}7sj4@RZ;Tv=rn``Kbp zYoQ>_isc&sfE=1iYo(X@*{U?M4HHKojSChiuu?Pu7$yI8#ER~JJ7RBbQRo4-Tmh_L z1>${9z5P@A2SG~{FOQ1LytkgIMimYS^z)#V4iSRDr_&3R+YNjqOLbrwS0V{cR{_Qj zeaG62qXbO>Pe~sLA>bT_z#776^D#EqSFpnDXqGYJ{v+!+S_Fi&8@5)>R z33(0~;`E9m5m~%{cJ>ghp88b}{e405Q@`e_L0X-+<8Ot73EGiBE@wO;IfbKRim7-S zrL$$!rR<>k7aJlAfub-w$!zf#^$$#M>-W37srF-L&L1$dLzCMyB6x@($|YbWbSV~xC}K}4{}g50K=jWadZE6mYdKUl%Nj1 zfWN{_wth{albL1BO6tecFvZnB02VI!;Wr*tACjkh$w8BDA1*2a@T%;nLg^MgP7at-Pf@n$GCiu^Wk7BBq{+#BW3@&-1do4cqd7solamI$96b z`yEx!^-_QXXDkR6i1v6=I1Y_1B6Q54F7+jA6PT?1j?|gn0(Pxl?Qi;k_@S78vFSH> zTYy%{XG!_YYUN+x*xbr`_9WtVM^JTN4MTYxW9fLMVh)n_YHy*qGIy6VFcC$^BT1}} zSZ1QR!!+R9qQyJ?la9HSrGg|SZK>y#3F0djIJ)&Q3*_3trN<&0Xg>NY-+EuRTv~~{ zNp}lc<(K*k`pkk+XOlIdo|E-|Jr>|knPu;F<^RcvKca8rdl8&7Exg0{vkL5jV-RTo z<=M-0KWimb@XI;n5U=b$mabY9P6i-BQBBFJBRFrJz%bmO`W$fKjTG{rG}4_3?*3yA z{#)Qfe6h*G-hC#!a=*A%X&>ERqjCJnyhi`97l2G3wZwL~}A9q6XUqJV^m@H6yHSn~bl>b7^sLQ|g3Z1Vveor?6i^$oK!Y zH8**?^OAtzD}4Az*}GhSqk-NX9CNJUtG8WHX2mT9$#Tp^_L)yQQN}5v$aT334x8LA zMJr~h+fcC}7zq*H56v{6@ERWMODBOypwUK+<59Y`ks&}|3IHIBEo5I|*eNCB$&&HT zkWq940kfL6lk8M6!$BayYP8%X0Q6nK`){e^d;pOS<8c^%f*f64$hZFerIB^UtNKiIcB%@vpL2<)?&^JTSZw@`!3p)c7sytqq}{5bq~}i zyJntdYavm-r2Hsxwd?oVbqjA8h#g#4Wl}kV;MDC3{!@eCJ&Kfr)IZopcO}caFr(J#Y*K4PuJF~3pVQSU+8l8bJ*C=Gg4V?gTKZjkA7c`Z&AX$;Mj#)^SbX@JA9 z5MQhuTf>qwbkjL+XFCBE2h5zlxj{^$tK<`N{7WqlQsiVXyM{bhX!W@glvga|z}m$F zKDz^d^8hwR+c0l|8%^WC;5npjplI2mS+EBM{Aqq>5*ZlU4B3PKq#V=eW+f`|x%9;`?F7|Mug!HQsi zb+ZPFV6PWASm>bW5Ie#n^!N~-gBHx`{4Yr>9fyu3*XX&q(vhx2xQmV1)pk342YxLb zy16pZ7(Ue$d8G*7^~TCcZ3|fWZ2h+?OxGcSoQrk~zsdaBIzszaa_|OzQ0U>rstb5cL3mXSE(f znOsmZL0I0l5?Pyr29%PP`gzHF#-5mGrOJ7)k(qdQ!8YXHq9Ag;0jMWfPTYj#BNB9> ziJ56Mm4NgjQ#2Dw=V?WQl-DJa>f_TirQ6JroHUtHf%QMweygc8bf`|tYe}3}nCVx8 zOPC=&lB#>Hj#g>0RR^R3C_52d)|{puLCx?*CF!7<0A2Coy94|w+M>*VFu|ApNwgYG zEg72&-5j?#XLA1Dc*qF7mv4gmv%_e`;iwa(Jyef16$5Tzb&0NL0l&QcdhJ**4Ms)F zsu>3klJ$F%(ypuKBtq51#8duof5G^MoOsEP_MX!)x6<|y5a83SjY3!$%Bz(Yr=2TH z2Vhet8|7t>y>Lv!Y!|+N-b;D>sTU!^|JLJnG*-pjYHoa_%&$Yi$d;UQa}bHuY`-=s z8TZ&7t}#h;W7%?-S^T8RQI4rYeA=`Wq~Qt00|Xmx#_{q)M6%?+T#nk}h!;?*2p}DT z)LG#-tR;yVv2N_0V3MjEjdki<#(V?ltMFq`3N7;IR0TTR{9>Jdvn-FPuN@n^aXXJ_ zCC{PKzRmg{lfUl9$Q-^PNv;E%3iZ3Cd_w2|AA!(EG0O_HR+5B$$JGOfVp`cUE~*Pm zLn8d7U^pF3Mur@mw#|LTJqGldf1X-?|^oB-%I0weY70Ar$ps-DKMPY$C9vwAHhQPynXl8utw6W8oy3bzjLqWSj zHq|ADts1vu^(vc{4ttg$`sB=xPyhiRewHWj8PQyv6W~umJav`0=g7ch?co1hyNzjZ z+c3Xa`ZUzl=qY#RHU?N+3d-qkqieZdMi7K`2U*!C@8KPPxc)CTe-tv!;|IY5*X`>; zGfo?JR-3m5zJ-jSEKbdpa2KjdewfY?OagB)v4;0ACZ41Mie0#0+fyJ#|GRcJ)!XX$ zvgpP0c9(0(XQZoi!s*pY_?|TgOVHBuc~>o9;FP(zjB6?mIaXcII_S&MDe}nGATEjA zrGq=W*uB4hrqZy}g$=2(TZZm7^VdDBe+%a&w8e5hO7G3|t(lkS*UwHhzp~Z~6i|U! z&mRRz)S|&(Nf-$(n3JK|PmU zC}uaoHKEa$5kYvH3+_$*6uy+7Y$r2&Wome@a=ws%cl?mjOL#}Qw%-uW9uGi%qLtcQ z$71o8{rGvW_6@sGPDzmXX!!dF=+n}@$2AAp_$IUIfdeicz7P{=@W0aKpikEkH+S_} zjUJCwL{uJX;H`dq+lX&#$G+S$akX%oXVw1AaT7TgpSU^GqeX)Gw1#ho13Mqm_hyr9 zizWeo`bE{B8O75Is4yB(ky(QJhf<<2 zb_0=6z2s$Y$}&5W!oE-} zG$Ox63-zRr-t5S?3ej*env%_3-H!%DHS1Y_mz$=1#gp`(=q2#ZJ){5q@Oit4?HUMy z!=!PY`lUR5WoR%&-bEz$+cMXXYhbB!j_>#ZUFGyn9sdDB?$h3QEAuCe!89>|H6|p) zXvBb2)yb%bTN2IEJQgv5;!VwhxnTSAit(tt>w;_hEtHB%m&Wm*MzLNp2R&W_9;`2a zO{REHR#~%rT4R1}OBg|K-gH&o<^R_JSDwu>KxGFG-%QA4QnQBy?chq%GU1^Q(fTY_ z1<7R)-fzn^XkJ60=et`V(|@=ux%Q4(wV+4vf;(KC864%J-CSw1dH$^VEi(>C6`1w& z@{rIQyK(P63!)Y+MG2Ju8VP8kn@FsGpDsja6YnFkP^iY`cmex?RB)Ux`ZKf!Uu!oTHEVt|g24s`dx&LBaCbv*QqimDBh&(;A`nkH@an$?~fBM^_O zR$V}1SN=;YTZkP&*Vus1HXM`#mc0eEo$Y50e4mU);WBjYblGuH3o*aq*$4zjm?LDO{KT)fNU?yvrv2Rv*8JwM@L^Zjk(D@w#MRaW zT#kCVgq<;u>t6xx^#6*P%s9pkVoHV^2j_cTavsNyK~5yE)0ta;fHTGxG1vvlIM3{R zCKC3v_?~e|D9NG$DDVXeMh@fdyv%1>>U5zTN9H66zu!o5D>P3=%*RzIVoMQD+UTdG z`6AT7NHy*3Z|&Kj@Iw&^TME+4=0xtKQiO;)SWhcUqePv)VWZOA9nR*%XnQeKZv&U5@y!%r?~ z?C|W4H^Y}kqvfcsnle(cvmg918nc{OR$)AdMjG5JF*@_#4KGKmgB}u-$Vj~{IX88I z%iDn|Lk?4art@jTAg2(x-u#S(KxhPvI9jPT1ciij$tv@ziAdlzZE&mI{*92cZi)mv zQ2oj5;F&jO-~!UU=a#y#ZnPFr?3aMNvT*i!tGquKKPc5^b{GEq{MR z&t|Oz_V1m(86-G8-Y8U5L%8cgryiUONt$M9Z2kwjkB5M7BuoNn1$_bQEi@lOIh6FI zRB!r1rm+1IYP@#s*h<9q-)S9lul)X}SzOzXg0NC-fa88KNObI*LGg-dc1&IcBQkUp z{}jf5`jY~bCD@5T$M}nGk6*uw&T76io&dT3${DKh_X%fx%rgi+k19TtS_C19Zt6rp z_*|9{kG^b#xMfvE?xhqwM97Wy)Z=EOWzsP&+hvMkAvW!d5KC+jvmEE0Tw&oww%A!4d5t7cMu(+pSFYMeK_^XN>S?o=ZdpxRbmU zRpF~(Mxr%QJ8;X>m2_!OU?%QiI?MQ{vgS?ov`ETzRDUs^Gb${y{O(ds>= zn*^!wq`1N#ISTF25zCygZ^TtqhB;9Lks?*Q!&Tm3#O;EoT7X#(Q^B#9*YesNJEcNs zitl{KCm@2tT)1Av<7VWj^RSvvtdKS)=m%p8o8pWD^l0Z6HHEQHs)C@`>$=H~7Fh5l zU|$2Wy)IwnHt_iFh6?3ZCyTD@-r6I7DcF0t4sfvWzi)1dG+EImt@Bn2I*ztqp0+xq zfE>`9UvGC0XJhb!=e|+Nnr^*tNDnsXGT^x<8;-*-1<7>*9tHm0x`E^(cS)Owz~gaNoFDZN zlu?Q_Vrb2DGENIfVMQ$57UR|Fe+G5R>tfh^VgFVcmrY5Ru$I~)(@&&QqWufeqjNI@ zOW+u_+2<)D0WGy_2TU)TI$eIQ-GacE_ND+Pj4hb0zBsx)6Om6qDFRo2mhJll`!Rb3 zmqd;V@kXHu8g4rMw(+as@ahDcvaZM?xQIpNQ3IrDtl36MF(3S!k!ZIf$}9$mgqP2G zo>Qg6Z|)Y#@U<$1{pUSB!(#~T^onkn>gaua`b4>eK;{d=XAwt*~xvq+ca)V%zz$=NaL@?5(oNoXYsLxM<}0-xu}k=4NR;V6WN)%V#* zHLk;-qpoJq`h!u4dUTd>x5PDK=tNYgPl>?}qXAn-@yn%brk)C(fvhNhp!;2;wFAJz zNOSHWUjBSCcRt~NGstMG3hx|AoIqD++*A*6e#{(HA~T@0ujABI%Gp;C^~oPq$H*Mv*?o;1u$p-}H!4Vg?^^P|mp{8Lt3;iimDD zJ@~EWKyF!b)q8~+CVTBPa)SSL-|kqzx7Fh1lEnd?K*e2ul8}Le;;>^)K_1`?t{VGv zQ=2IFWzs-1u0>RJx1wudYp^}!GyS4)Jf<|x0S^CHiTdHoZH14TZ5o5lM$0o=y0t)> zl_eO-BqD|56&wr@7JzZ28u;3WalS5#Bd^ArZS*3`3|Vwu3~npK1rHo3e+z^2O@<>l z#j~xewxAh*iu&h3t-aRcydvh_nTjl)rk2iTrD?!QydwK^X6=>bbtkH^vRZLcF|MVW-uL30#TFjrxSn`Wg_|$gQ)`Jqs zWh+@+HtmO4d$27W_@suYn%1?l-bpSfV6v+ZLUKWW_w{a`U%JeT&>gR+sX zW=g_FunCvJX{5&0krlf>sfS}E_z{`|O*%mY+E=pP`lKnR#@v>K2n_;IU#WRgK6ab4 z+8s|a_H>$g);+sPnUA6ICK$2f z2QE>s#k5$xu>@dBaZzFA#ihYfY3JU5g5#M=$?UoloD?%t0gtL19~!yES~Y-WR8|uH zF!zEZf4p26D+h0oBQMd2e7Cm0$n3zK15ZxZ2to+N&Rkv)$bSmk_UDd z_L;KD@L}27obo56nRoy+H&3NIc@B-koUJ?56Rp(_iveN6ViOm6-ol=?5#@q^#ZjD> zL*nM0*L8B}O(i2mPgfm?-F_);z}-xAgg!un96oYww3Cv&_r5qTj6ZaNEQP8D?96*E zW+v`432$EjcNdxUzS~nFGw|a4X=mES)OsPKnj$O2+9w0?q=J@V70N&?6#)6*B1=qz zf}lE$TGXbca4?IF3(_>f$g78ctbRNC^*5lAsmAXc)h>sKCf2C+xm8AtA~kK}2hJyp z4&L;!TSN3RxL^-!r4y z8{K=pgxeobYQU9nykdC)FZ>esw9#3ldEX+t<`)j7c^OjX(2px8@lNmX34{Y? z*m<{#O?6p}h4!%gr-PBKHKrj0G6M}D?aHZR;N|i%J6e+nh=~bRI^64gfKn-lmfq(9 z8`ax?k3o?B6B=B!q?6hlG{Qh@51BBDV#1+m|6wHo26hafB4|kKQYwpmqM8rcAeIqN zY&?g$%+nO}v3v0o)4N}P-kkZ1EgdWsS|oLGwGVj%AbCm~KB05bZiU$1=6-hHgMrtv2w~pcnq0Fdj;fP8Bv0001nb0zQq0j84&+wMESu1ohmFb#_W000000a;qK^R{aM diff --git a/data-raw/_pre_commit_checks.R b/data-raw/_pre_commit_checks.R index dc2c39ffc..fccb6b7c2 100644 --- a/data-raw/_pre_commit_checks.R +++ b/data-raw/_pre_commit_checks.R @@ -207,17 +207,24 @@ pre_commit_lst$MO_RELEVANT_GENERA <- c( "Angiostrongylus", "Anisakis", "Anopheles", + "Apiotrichum", "Apophysomyces", "Arthroderma", + "Ascaris", "Aspergillus", "Aureobasidium", + "Babesia", + "Balamuthia", + "Balantioides", "Basidiobolus", "Beauveria", "Bipolaris", "Blastobotrys", "Blastocystis", "Blastomyces", + "Brugia", "Candida", + "Candidozyma", "Capillaria", "Chaetomium", "Chilomastix", @@ -226,6 +233,7 @@ pre_commit_lst$MO_RELEVANT_GENERA <- c( "Cladophialophora", "Cladosporium", "Clavispora", + "Clonorchis", "Coccidioides", "Cokeromyces", "Conidiobolus", @@ -236,21 +244,31 @@ pre_commit_lst$MO_RELEVANT_GENERA <- c( "Cryptosporidium", "Cunninghamella", "Curvularia", + "Cutaneotrichosporon", "Cyberlindnera", + "Cyclospora", + "Cystoisospora", "Debaryozyma", "Demodex", "Dermatobia", "Dientamoeba", "Diphyllobothrium", "Dirofilaria", + "Diutina", + "Echinococcus", "Echinostoma", + "Emergomyces", + "Emmonsia", + "Encephalitozoon", "Entamoeba", "Enterobius", + "Enterocytozoon", "Epidermophyton", "Exidia", "Exophiala", "Exserohilum", "Fasciola", + "Fasciolopsis", "Fonsecaea", "Fusarium", "Geotrichum", @@ -268,17 +286,21 @@ pre_commit_lst$MO_RELEVANT_GENERA <- c( "Hymenolepis", "Hypomyces", "Hysterothylacium", + "Kazachstania", "Kloeckera", "Kluyveromyces", "Kodamaea", "Lacazia", "Leishmania", "Lichtheimia", + "Loa", "Lodderomyces", "Lomentospora", "Madurella", + "Magnusiomyces", "Malassezia", "Malbranchea", + "Mansonella", "Metagonimus", "Meyerozyma", "Microascus", @@ -288,15 +310,20 @@ pre_commit_lst$MO_RELEVANT_GENERA <- c( "Mortierella", "Mucor", "Mycocentrospora", + "Naegleria", + "Nakaseomyces", "Nannizzia", "Necator", "Nectria", + "Neocosmospora", "Ochroconis", "Oesophagostomum", "Oidiodendron", + "Onchocerca", "Opisthorchis", "Paecilomyces", "Paracoccidioides", + "Paragonimus", "Pediculus", "Penicillium", "Phaeoacremonium", @@ -316,6 +343,7 @@ pre_commit_lst$MO_RELEVANT_GENERA <- c( "Pulex", "Purpureocillium", "Quambalaria", + "Rasamsonia", "Rhinocladiella", "Rhizomucor", "Rhizopus", @@ -323,6 +351,7 @@ pre_commit_lst$MO_RELEVANT_GENERA <- c( "Saccharomyces", "Saksenaea", "Saprochaete", + "Sarcocystis", "Sarcoptes", "Scedosporium", "Schistosoma", @@ -344,6 +373,7 @@ pre_commit_lst$MO_RELEVANT_GENERA <- c( "Talaromyces", "Teleomorph", "Toxocara", + "Toxoplasma", "Trichinella", "Trichobilharzia", "Trichoderma", diff --git a/data-raw/_reproduction_scripts/reproduction_of_microorganisms.R b/data-raw/_reproduction_scripts/reproduction_of_microorganisms.R index b5bf2578c..79fa96e30 100644 --- a/data-raw/_reproduction_scripts/reproduction_of_microorganisms.R +++ b/data-raw/_reproduction_scripts/reproduction_of_microorganisms.R @@ -28,10 +28,6 @@ # ==================================================================== # -# TODO Giardia zat in protozoa en animalia, dus sorteren op prioriteit kingdom en dan distinct op fullname - - - # ! THIS SCRIPT REQUIRES AT LEAST 16 GB RAM ! # (at least 12 GB will be used by the R session for the size of the files) @@ -57,9 +53,11 @@ # 5. Go to BacDive data base for the oxygen tolerance and cell shape. # Go to https://bacdive.dsmz.de/advsearch, filter 'Oxygen tolerance' or # 'Cell shape' on "*" and click Submit and click on the 'Download tabel as CSV' button -# 6. Set these locations to the paths where the files are: -folder_location <- "~/Downloads/" -file_gbif <- paste0(folder_location, "/xr_latest_dwca/Taxon.tsv") +# 6. Set these locations to the paths where the files are (or set the environment variable +# AMR_TAXONOMY_SOURCE_DIR to the folder, e.g. for an unattended run): +folder_location <- Sys.getenv("AMR_TAXONOMY_SOURCE_DIR", unset = "~/Downloads/") +folder_location <- paste0(sub("/+$", "", path.expand(folder_location)), "/") +file_gbif <- paste0(folder_location, "xr_latest_dwca/Taxon.tsv") file_lpsn <- paste0(folder_location, "taxonomy.csv") file_mycobank <- paste0(folder_location, "MBList-2.xlsx") @@ -68,26 +66,43 @@ file_bartlett <- "data-raw/bartlett_et_al_2022_human_pathogens.xlsx" file_oxygen_tolerance <- paste0(folder_location, "oxygen.csv") file_cell_shape <- paste0(folder_location, "shape.csv") -# 4. Run the rest of this script line by line and check everything :) +# 7. Run the rest of this script line by line and check everything :) +# All review moments use review(): in an interactive session they open View(), otherwise they +# print the first rows (50, or set options(AMR_build_print_rows = ...)). Set +# options(AMR_build_view = FALSE) to only print, e.g. for a run with source(..., echo = TRUE) +# and reading the log afterwards. For an unattended run by Claude, see +# data-raw/_reproduction_scripts/reproduction_of_microorganisms_CLAUDE.md. -if (!file.exists(file_gbif)) { - stop("GBIF file not found") +# The source files can only be downloaded by a human (registration, manual export buttons), so this +# script stops immediately if any of them is missing, instead of continuing with old data. +raw_files <- c( + "GBIF/COL Taxon.tsv (step 1)" = file_gbif, + "LPSN taxonomy.csv (step 2)" = file_lpsn, + "MycoBank MBList-2.xlsx (step 3)" = file_mycobank, + "Bartlett et al. Excel file (step 4)" = file_bartlett, + "BacDive oxygen.csv (step 5)" = file_oxygen_tolerance, + "BacDive shape.csv (step 5)" = file_cell_shape +) +raw_missing <- raw_files[!file.exists(raw_files)] +if (length(raw_missing) > 0) { + stop( + "The following source files are missing and must be downloaded by a human first ", + "(see steps 1-6 at the top of this script):\n", + paste0(" - ", names(raw_missing), ": ", raw_missing, collapse = "\n"), + call. = FALSE + ) } -if (!file.exists(file_lpsn)) { - stop("LPSN file not found") -} -if (!file.exists(file_mycobank)) { - stop("MB file not found") -} -if (!file.exists(file_bartlett)) { - stop("Bartlett et al. Excel file not found") -} -if (!file.exists(file_oxygen_tolerance)) { - stop("BacDive oxygen tolerance CSV file not found") -} -if (!file.exists(file_cell_shape)) { - stop("BacDive cell shape CSV file not found") +# downloaded files of more than 120 days old are probably from the previous update +raw_downloads <- raw_files[raw_files != file_bartlett] +raw_age <- as.numeric(Sys.Date() - as.Date(file.mtime(raw_downloads)), units = "days") +if (any(raw_age > 120)) { + warning( + "These source files are older than 120 days, check that they were downloaded for this update: ", + toString(names(raw_downloads)[raw_age > 120]), + call. = FALSE + ) } +rm(raw_missing, raw_downloads, raw_age) library(dplyr) library(tidyr) @@ -97,9 +112,47 @@ library(progress) # to show progress bars library(readxl) # to read the MycoBank and Bartlett Excel files devtools::load_all(".") # to load the AMR package +# keep the previous data set at hand, `AMR::microorganisms` will be replaced at the end of this script +microorganisms_old <- AMR::microorganisms + # Helper functions -------------------------------------------------------------------------------- +# show data that must be reviewed, without blocking a non-interactive run +# with options(AMR_build_review_dir = "some/folder"), every review is also saved as a CSV file in that +# folder, so that a human can review them afterwards (e.g. in a pull request). The file name is based on +# the title only, so running a section again overwrites its own file. +review <- function(x, title = deparse(substitute(x))) { + message("\n>>> REVIEW: ", title, " (", nrow(x), " rows)") + review_dir <- getOption("AMR_build_review_dir") + if (!is.null(review_dir)) { + dir.create(review_dir, recursive = TRUE, showWarnings = FALSE) + file_name <- paste0(gsub("[^a-z0-9]+", "_", gsub("^[^a-z0-9]+|[^a-z0-9]+$", "", tolower(title))), ".csv") + out <- as.data.frame(x) + # list columns (such as `snomed`) cannot be written to CSV + out[] <- lapply(out, function(col) if (is.list(col)) vapply(col, toString, character(1)) else col) + utils::write.csv(out, file.path(review_dir, file_name), row.names = FALSE, na = "") + } + if (nrow(x) > 0) { + if (interactive() && isTRUE(getOption("AMR_build_view", TRUE))) { + utils::View(x, title) + } else { + print(utils::head(as.data.frame(x), getOption("AMR_build_print_rows", 50))) + } + } + invisible(x) +} + +# the priority of sources, used everywhere in this script +source_priority <- c("LPSN" = 1L, "MycoBank" = 2L, "GBIF" = 3L, "manually added" = 4L, "inferred" = 5L) +source_prio <- function(source) { + coalesce(unname(source_priority[source]), 6L) +} + +# 'inferred' is used during this script for records that are not in any source, but that we must +# create ourselves (e.g. a genus of which only species are available). These are NOT manually curated +# and must therefore follow all the relevance filters. They are renamed to "manually added" at the end. + get_author_year <- function(ref) { # Only keep first author, e.g. transform 'Smith, Jones, 2011' to 'Smith et al., 2011' @@ -119,7 +172,7 @@ get_author_year <- function(ref) { authors2 <- trimws(gsub("^[(](.*)[)]$", "\\1", authors2)) # only take part after brackets if there's a name authors2 <- if_else( - grepl(".*[)] [a-zA-Z]+.*", authors2), + authors2 %like_case% ".*[)] [a-zA-Z]+.*", gsub(".*[)] (.*)", "\\1", authors2), authors2 ) @@ -186,14 +239,14 @@ get_author_year <- function(ref) { ref <- gsub("'", "", ref) # a lot start with a lowercase character - fix that - ref[!grepl("^d[A-Z]", ref)] <- gsub( + ref[ref %unlike_case% "^d[A-Z]"] <- gsub( "^([a-z])", "\\U\\1", - ref[!grepl("^d[A-Z]", ref)], + ref[ref %unlike_case% "^d[A-Z]"], perl = TRUE ) # specific one for the French that are named dOrbigny - ref[grepl("^d[A-Z]", ref)] <- gsub("^d", "d'", ref[grepl("^d[A-Z]", ref)]) + ref[ref %like_case% "^d[A-Z]"] <- gsub("^d", "d'", ref[ref %like_case% "^d[A-Z]"]) ref <- gsub(" +", " ", ref) ref <- trimws(ref) ref <- gsub("^NA, ?", "", ref) @@ -203,36 +256,71 @@ get_author_year <- function(ref) { # to retrieve LPSN and authors from LPSN website # e.g., get_lpsn_and_author("genus", "Klebsiella") -get_lpsn_and_author <- function(rank, name) { +# results are cached in `lpsn_cache` (and in data-raw/lpsn_scrape_cache.rds using save_lpsn_cache()), +# so an interrupted run does not have to download thousands of pages again +lpsn_cache_file <- "data-raw/lpsn_scrape_cache.rds" +lpsn_cache <- new.env() +if (file.exists(lpsn_cache_file)) { + # only use a cache of this month, the LPSN is updated continuously + if (format(file.mtime(lpsn_cache_file), "%Y-%m") == format(Sys.Date(), "%Y-%m")) { + list2env(readRDS(lpsn_cache_file), envir = lpsn_cache) + } +} +save_lpsn_cache <- function() { + saveRDS(as.list(lpsn_cache), lpsn_cache_file, version = 2) +} +get_lpsn_and_author <- function(rank, name, tries = 3) { name <- gsub("^Candidatus ", "", name) url <- paste0( "https://lpsn.dsmz.de/", tolower(rank), "/", gsub(" ", "-", tolower(name)) ) - page_txt <- tryCatch({ - resp <- curl::curl_fetch_memory(url) - if (resp$status_code == 200) { - read_html(rawToChar(resp$content)) - } else { - NULL + if (!is.null(lpsn_cache[[url]])) { + return(lpsn_cache[[url]]) + } + page_txt <- NULL + not_found <- FALSE + for (try in seq_len(tries)) { + # be polite to the LPSN server + Sys.sleep(0.1) + page_txt <- tryCatch({ + resp <- curl::curl_fetch_memory(url) + if (resp$status_code == 200) { + read_html(rawToChar(resp$content)) + } else { + # 404 is a valid answer, other codes (e.g. 429, 503) are worth a retry + not_found <- resp$status_code == 404 + NULL + } + }, error = function(e) NULL) + if (!is.null(page_txt) || not_found) { + break } - }, error = function(e) NULL) - if (is.null(page_txt)) { - warning("No LPSN found for ", tolower(rank), " '", name, "'") - lpsn <- NA_character_ - ref <- NA_character_ - status <- "unknown" - } else { + Sys.sleep(5 * try) + } + if (!is.null(page_txt)) { page_txt <- page_txt %>% html_element("#detail-page") %>% html_text() + } + # a page without a record number is no valid LPSN record (otherwise the whole page text would be + # returned by gsub() below and stored as the LPSN ID) + if (is.null(page_txt) || is.na(page_txt) || page_txt %unlike_case% "Record number:") { + if (!not_found) { + warning("No LPSN found for ", tolower(rank), " '", name, "'", call. = FALSE) + } + out <- c("lpsn" = NA_character_, "ref" = NA_character_, "status" = "unknown") + } else { lpsn <- gsub( ".*Record number:[\r\n\t ]*([0-9]+).*", "\\1", page_txt, perl = FALSE ) + if (lpsn %unlike_case% "^[0-9]+$") { + lpsn <- NA_character_ + } ref <- page_txt %>% gsub(".*?Name: (.*[0-9]{4}?).*", "\\1", ., perl = FALSE) %>% gsub(name, "", ., fixed = TRUE) %>% @@ -253,8 +341,41 @@ get_lpsn_and_author <- function(rank, name) { } else { status <- "not validly published" } + out <- c("lpsn" = lpsn, "ref" = ref, "status" = status) } - c("lpsn" = lpsn, "ref" = ref, "status" = status) + # do not cache network failures, so they will be retried in a next run + if (!is.null(page_txt) || not_found) { + lpsn_cache[[url]] <- out + } + out +} + +# get the LPSN record of all unique taxa of a certain rank (with their higher taxonomy), +# used for families until domains below. Binding once at the end is much faster than binding in the loop. +get_lpsn_rank <- function(taxonomy_lpsn, rank) { + higher_ranks <- c("domain", "kingdom", "phylum", "class", "order", "family") + higher_ranks <- higher_ranks[seq_len(which(higher_ranks == rank) - 1)] + taxa <- taxonomy_lpsn %>% + filter(!is.na(.data[[rank]])) %>% + distinct(.data[[rank]], .keep_all = TRUE) %>% + select(all_of(c(higher_ranks, rank))) + pb <- progress_bar$new( + total = nrow(taxa), + format = paste0(rank, " [:bar] :current/:total :eta") + ) + info <- lapply(taxa[[rank]], function(nm) { + pb$tick() + get_lpsn_and_author(tools::toTitleCase(rank), nm) + }) + save_lpsn_cache() + taxa %>% + mutate( + rank = rank, + status = vapply(info, function(x) unname(x["status"]), character(1)), + source = "LPSN", + lpsn = vapply(info, function(x) unname(x["lpsn"]), character(1)), + ref = vapply(info, function(x) unname(x["ref"]), character(1)) + ) } # this will e.g. take the family from the root genus record, and gives all species of that family @@ -269,26 +390,187 @@ get_top_lvl <- function(current, rank, source, rank_target, target) { } else if (!rank_target %in% rank) { current.bak[1] } else { + pick <- function(current, rank) { + out <- current[rank == rank_target][1] + if (out %in% c("", NA)) { + out <- names(sort( + table(current[which(!current %in% c("", NA))]), + decreasing = TRUE + )[1]) + if (is.null(out)) { + out <- "" + } + } + out + } + out <- "" if (n_distinct(source) > 1 && "GBIF" %in% source) { # prefer LPSN and MycoBank over GBIF, which is often not up-to-date at all - current <- current[source != "GBIF"] - rank <- rank[source != "GBIF"] - source <- source[source != "GBIF"] + out <- pick(current[source != "GBIF"], rank[source != "GBIF"]) } - out <- current[rank == rank_target][1] - if (out %in% c("", NA)) { - out <- names(sort( - table(current[which(!current %in% c("", NA))]), - decreasing = TRUE - )[1]) - if (is.null(out)) { - out <- "" - } + if (out == "") { + # but never return an empty value if GBIF does have one (e.g. Mucor has no family in MycoBank) + out <- pick(current, rank) } out } } +# for a set of records, take the most reliable non-empty value: of the record with the own rank +# (e.g. the family record itself when determining the order of a family), else of the most +# authoritative source, else the most common value +consensus_value <- function(x, rank, source, own_rank) { + ok <- !x %in% c("", NA) + if (!any(ok)) { + return("") + } + x <- x[ok] + rank <- rank[ok] + source <- source[ok] + own <- x[rank == own_rank] + if (length(own) > 0) { + return(own[order(source_prio(source[rank == own_rank]))][1]) + } + best_source <- min(source_prio(source)) + x <- x[source_prio(source) == best_source] + names(sort(table(x), decreasing = TRUE))[1] +} + +# all records whose fullname is not in `taxonomy` yet, but which are referred to by other records +# (e.g. a family of which only genera are available) - these will be added as 'inferred' records, +# with identifiers from GBIF where available +add_missing_parents <- function(taxonomy, current_gbif) { + rank_levels <- c("domain", "kingdom", "phylum", "class", "order", "family", "genus") + missing_higher <- bind_rows(lapply(rank_levels, function(rank_name) { + candidates <- taxonomy %>% + filter(.data[[rank_name]] != "") %>% + distinct(across(all_of(unique(c("domain", rank_levels[seq_len(which(rank_levels == rank_name))]))))) %>% + mutate(fullname = .data[[rank_name]], rank = rank_name) %>% + anti_join( + taxonomy %>% filter(rank == rank_name), + by = unique(c("domain", setNames(rank_name, rank_name), "rank")) + ) %>% + # one record per taxon, inconsistent higher taxonomy will be fixed later + distinct(domain, fullname, .keep_all = TRUE) + if (nrow(candidates) == 0) { + return(tibble()) + } + message("Adding missing: ", rank_name, "... n = ", nrow(candidates)) + # enrich from GBIF + gbif_lookup <- current_gbif %>% + filter(taxonRank == rank_name) %>% + select( + all_of(unique(c("domain", rank_name))), + ref = scientificNameAuthorship, + gbif = taxonID, + gbif_parent = parentNameUsageID + ) %>% + distinct(across(all_of(unique(c("domain", rank_name)))), .keep_all = TRUE) + candidates %>% + left_join(gbif_lookup, by = unique(c("domain", rank_name))) %>% + mutate( + source = if_else(!is.na(gbif), "GBIF", "inferred"), + status = if_else(!is.na(gbif), "accepted", "unknown") + ) + })) + # species implied by subspecies but missing as a species-rank row + missing_species <- taxonomy %>% + filter(species != "") %>% + distinct(domain, genus, species, .keep_all = TRUE) %>% + select(domain:species) %>% + mutate(fullname = paste(genus, species), rank = "species") %>% + anti_join( + taxonomy %>% filter(rank == "species"), + by = c("domain", "genus", "species", "rank") + ) %>% + left_join( + current_gbif %>% + filter(taxonRank == "species") %>% + select( + domain, genus, + species = specificEpithet, + ref = scientificNameAuthorship, + gbif = taxonID, + gbif_parent = parentNameUsageID + ) %>% + distinct(domain, genus, species, .keep_all = TRUE), + by = c("domain", "genus", "species") + ) %>% + mutate( + source = if_else(!is.na(gbif), "GBIF", "inferred"), + status = if_else(!is.na(gbif), "accepted", "unknown") + ) + if (nrow(missing_species) > 0) { + message("Adding missing: species... n = ", nrow(missing_species)) + } + taxonomy %>% + bind_rows(missing_higher, missing_species) %>% + mutate( + domain = if_else(is.na(domain) | domain == "", kingdom, domain), + across(kingdom:subspecies, \(x) if_else(is.na(x), "", x)) + ) +} + +# Clinically relevant genera ---------------------------------------------------------------------- + +# One set of genera that must never get lost, used by all filters below. It contains: +# - our own curated list of non-bacterial genera (MO_RELEVANT_GENERA, see data-raw/_pre_commit_checks.R) +# - the WHO priority genera +# - all genera of the human pathogens in Bartlett et al. (2022) +# - all genera that are used in the other data sets of this package (breakpoints, intrinsic resistance, etc.) +# - all genera that were clinically relevant in the previous version of this data set +# - and of all these: their current names and their synonyms (e.g. Candida -> Candidozyma, Nakaseomyces) +pathogens <- read_excel(file_bartlett, sheet = "Tab 6 Full List") + +genera_of_mo <- function(x) { + unique(microorganisms_old$genus[match(as.character(x), as.character(microorganisms_old$mo))]) +} +relevant_genera <- c( + AMR:::MO_RELEVANT_GENERA, + AMR:::MO_WHO_PRIORITY_GENERA, + pathogens$genus, + genera_of_mo(AMR::clinical_breakpoints$mo), + genera_of_mo(AMR::intrinsic_resistant$mo), + genera_of_mo(AMR::microorganisms.groups$mo), + genera_of_mo(AMR::microorganisms.groups$mo_group), + genera_of_mo(AMR::microorganisms.codes$mo), + genera_of_mo(AMR::example_isolates$mo), + microorganisms_old$genus[microorganisms_old$prevalence < 2 & microorganisms_old$domain != "Bacteria"] +) +relevant_genera <- relevant_genera[!relevant_genera %in% c("", NA) & relevant_genera %unlike% "unknown"] +relevant_species <- microorganisms_old %>% + filter(genus %in% relevant_genera, rank %in% c("species", "subspecies")) %>% + pull(fullname) +relevant_genera <- c( + relevant_genera, + # current names of all their species, e.g. Candida auris -> Candidozyma auris + mo_genus(unique(unlist(mo_current(relevant_species))), keep_synonyms = TRUE), + # synonyms of the genera themselves + unlist(mo_synonyms(unique(relevant_genera), keep_synonyms = FALSE)) +) %>% + strsplit(" ", fixed = TRUE) %>% + vapply(function(x) x[1], character(1)) %>% + unique() %>% + sort() +relevant_genera <- relevant_genera[!relevant_genera %in% c("", NA) & relevant_genera %unlike% "unknown"] +message(length(relevant_genera), " clinically relevant genera will be protected") +rm(relevant_species) + +# Genus names that exist in multiple domains (homonyms) cannot always be resolved automatically. +# E.g., Capillaria (nematode vs fungus) and Necator (hookworm vs fungus) would otherwise end up as +# fungi, since Fungi have priority over Animalia. These overrides have the highest priority. +# A report of all unresolved homonyms of relevant genera is shown below in 'Combine the datasets', +# so extend this list there when needed. +genus_domain_override <- c( + "Capillaria" = "Animalia", + "Graphium" = "Fungi", + "Hymenolepis" = "Animalia", + "Necator" = "Animalia", + "Schizophyllum" = "Fungi", + "Trichophyton" = "Fungi", + "Wangiella" = "Fungi" +) + # MB/ June 2024: after years still useless, does not contain full taxonomy, e.g. LPSN::request(cred, category = "family") is empty. # get_from_lpsn <- function (user, pw) { # if (!"LPSN" %in% rownames(utils::installed.packages())) { @@ -406,176 +688,61 @@ taxonomy_lpsn_missing <- taxonomy_lpsn_missing %>% distinct() saveRDS(taxonomy_lpsn_missing, "data-raw/taxonomy_lpsn_missing.rds") # taxonomy_lpsn_missing <- readRDS("data-raw/taxonomy_lpsn_missing.rds") -# duplicate genera: -taxonomy_lpsn_missing %>% filter(genus %in% taxonomy_lpsn_missing$genus[duplicated(taxonomy_lpsn_missing$genus)]) +# duplicate genera (homonyms) would duplicate rows in the join below: +review( + taxonomy_lpsn_missing %>% filter(genus %in% taxonomy_lpsn_missing$genus[duplicated(taxonomy_lpsn_missing$genus)]), + "LPSN genera with multiple lineages" +) # look them up on LPSN, then find out which to keep (the ones validly published under ICNP) # to remove: taxonomy_lpsn_missing <- taxonomy_lpsn_missing %>% - filter(!(genus == "Halalkalibacterium" & family == "Balneolaceae"), - !(genus == "Pusillimonas" & family == "Oscillospiraceae"), - !(genus == "Rhodococcus" & family == "Chroococcaceae")) + filter( + !(genus == "Halalkalibacterium" & family == "Balneolaceae"), + !(genus == "Pusillimonas" & family == "Oscillospiraceae"), + !(genus == "Rhodococcus" & family == "Chroococcaceae") + ) +# any remaining duplicates: keep the most complete lineage, so that the join below cannot duplicate records +if (anyDuplicated(taxonomy_lpsn_missing$genus) > 0) { + warning( + "Genera with multiple LPSN lineages left, keeping the most complete one: ", + toString(unique(taxonomy_lpsn_missing$genus[duplicated(taxonomy_lpsn_missing$genus)])), + call. = FALSE + ) + taxonomy_lpsn_missing <- taxonomy_lpsn_missing %>% + arrange(genus, rowSums(is.na(across(c(domain, kingdom, phylum, class, order, family))))) %>% + distinct(genus, .keep_all = TRUE) +} taxonomy_lpsn <- taxonomy_lpsn %>% - left_join(taxonomy_lpsn_missing, by = "genus") %>% - select(domain:family, everything()) %>% + left_join(taxonomy_lpsn_missing, by = "genus", relationship = "many-to-one") %>% + select(domain, kingdom, phylum, class, order, family, everything()) %>% # remove entries like "[Bacteria, no family]" and "[Bacteria, no class]" - mutate_all(function(x) if_else(x %like_case% " no ", NA_character_, x)) + mutate(across( + c(domain, kingdom, phylum, class, order, family), + function(x) if_else(x %like_case% " no ", NA_character_, x) + )) taxonomy_lpsn.bak2 <- taxonomy_lpsn -# download family taxonomic data (e.g. authors of Enterobacteriaceae) directly from LPSN website using scraping, by using get_lpsn_and_author() -# try it first: +# download higher taxonomic data (e.g. authors of Enterobacteriaceae) directly from LPSN website using +# scraping, by using get_lpsn_and_author(). Try it first: # get_lpsn_and_author("genus", "Escherichia") # get_lpsn_and_author("family", "Enterobacteriaceae") -pb <- progress_bar$new( - total = length(unique(taxonomy_lpsn$family)), - format = "[:bar] :current/:total :eta" -) -for (f in unique(taxonomy_lpsn$family)) { - pb$tick() - if (is.na(f)) { - next - } - tax_info <- get_lpsn_and_author("Family", f) - taxonomy_lpsn <- taxonomy_lpsn %>% - bind_rows(tibble( - domain = taxonomy_lpsn$domain[which(taxonomy_lpsn$family == f)[1]], - kingdom = taxonomy_lpsn$kingdom[which(taxonomy_lpsn$family == f)[1]], - phylum = taxonomy_lpsn$phylum[which(taxonomy_lpsn$family == f)[1]], - class = taxonomy_lpsn$class[which(taxonomy_lpsn$family == f)[1]], - order = taxonomy_lpsn$order[which(taxonomy_lpsn$family == f)[1]], - family = f, - rank = "family", - status = unname(tax_info["status"]), - source = "LPSN", - lpsn = unname(tax_info["lpsn"]), - ref = unname(tax_info["ref"]) - )) -} -# download order directly from LPSN website using scraping -# try it first: # get_lpsn_and_author("order", "Enterobacterales") -pb <- progress_bar$new( - total = length(unique(taxonomy_lpsn$order)), - format = "[:bar] :current/:total :eta" -) -for (o in unique(taxonomy_lpsn$order)) { - pb$tick() - if (is.na(o)) { - next - } - tax_info <- get_lpsn_and_author("Order", o) - taxonomy_lpsn <- taxonomy_lpsn %>% - bind_rows(tibble( - domain = taxonomy_lpsn$domain[which(taxonomy_lpsn$order == o)[1]], - kingdom = taxonomy_lpsn$kingdom[which(taxonomy_lpsn$order == o)[1]], - phylum = taxonomy_lpsn$phylum[which(taxonomy_lpsn$order == o)[1]], - class = taxonomy_lpsn$class[which(taxonomy_lpsn$order == o)[1]], - order = o, - rank = "order", - status = unname(tax_info["status"]), - source = "LPSN", - lpsn = unname(tax_info["lpsn"]), - ref = unname(tax_info["ref"]) - )) -} -# download class directly from LPSN website using scraping -# try it first: # get_lpsn_and_author("class", "Gammaproteobacteria") -pb <- progress_bar$new( - total = length(unique(taxonomy_lpsn$class)), - format = "[:bar] :current/:total :eta" -) -for (cc in unique(taxonomy_lpsn$class)) { - pb$tick() - if (is.na(cc)) { - next - } - tax_info <- get_lpsn_and_author("Class", cc) - taxonomy_lpsn <- taxonomy_lpsn %>% - bind_rows(tibble( - domain = taxonomy_lpsn$domain[which(taxonomy_lpsn$class == cc)[1]], - kingdom = taxonomy_lpsn$kingdom[which(taxonomy_lpsn$class == cc)[1]], - phylum = taxonomy_lpsn$phylum[which(taxonomy_lpsn$class == cc)[1]], - class = cc, - rank = "class", - status = unname(tax_info["status"]), - source = "LPSN", - lpsn = unname(tax_info["lpsn"]), - ref = unname(tax_info["ref"]) - )) -} -# download phylum directly from LPSN website using scraping -# try it first: # get_lpsn_and_author("phylum", "Pseudomonadota") -pb <- progress_bar$new( - total = length(unique(taxonomy_lpsn$phylum)), - format = "[:bar] :current/:total :eta" -) -for (p in unique(taxonomy_lpsn$phylum)) { - pb$tick() - if (is.na(p)) { - next - } - tax_info <- get_lpsn_and_author("Phylum", p) - taxonomy_lpsn <- taxonomy_lpsn %>% - bind_rows(tibble( - domain = taxonomy_lpsn$domain[which(taxonomy_lpsn$phylum == p)[1]], - kingdom = taxonomy_lpsn$kingdom[which(taxonomy_lpsn$phylum == p)[1]], - phylum = p, - rank = "phylum", - status = unname(tax_info["status"]), - source = "LPSN", - lpsn = unname(tax_info["lpsn"]), - ref = unname(tax_info["ref"]) - )) -} -# download kingdom directly from LPSN website using scraping -# try it first: # get_lpsn_and_author("kingdom", "Pseudomonadati") -pb <- progress_bar$new( - total = length(unique(taxonomy_lpsn$kingdom)), - format = "[:bar] :current/:total :eta" -) -for (k in unique(taxonomy_lpsn$kingdom)) { - pb$tick() - if (is.na(k)) { - next - } - tax_info <- get_lpsn_and_author("Kingdom", k) - taxonomy_lpsn <- taxonomy_lpsn %>% - bind_rows(tibble( - domain = taxonomy_lpsn$domain[which(taxonomy_lpsn$kingdom == k)[1]], - kingdom = k, - rank = "kingdom", - status = unname(tax_info["status"]), - source = "LPSN", - lpsn = unname(tax_info["lpsn"]), - ref = unname(tax_info["ref"]) - )) -} -# download domain directly from LPSN website using scraping -# try it first: # get_lpsn_and_author("domain", "Bacteria") -pb <- progress_bar$new( - total = length(unique(taxonomy_lpsn$domain)), - format = "[:bar] :current/:total :eta" -) -for (d in unique(taxonomy_lpsn$domain)) { - pb$tick() - if (is.na(d)) { - next - } - tax_info <- get_lpsn_and_author("Domain", d) - taxonomy_lpsn <- taxonomy_lpsn %>% - bind_rows(tibble( - domain = d, - rank = "domain", - status = unname(tax_info["status"]), - source = "LPSN", - lpsn = unname(tax_info["lpsn"]), - ref = unname(tax_info["ref"]) - )) -} +taxonomy_lpsn <- taxonomy_lpsn %>% + bind_rows( + get_lpsn_rank(taxonomy_lpsn, "family"), + get_lpsn_rank(taxonomy_lpsn, "order"), + get_lpsn_rank(taxonomy_lpsn, "class"), + get_lpsn_rank(taxonomy_lpsn, "phylum"), + get_lpsn_rank(taxonomy_lpsn, "kingdom"), + get_lpsn_rank(taxonomy_lpsn, "domain") + ) +# these could not be downloaded at all (network errors), run the lines above again to retry them: +review(taxonomy_lpsn %>% filter(status == "unknown"), "LPSN higher taxa that could not be downloaded") taxonomy_lpsn <- taxonomy_lpsn %>% filter(status != "not validly published") @@ -627,7 +794,12 @@ taxonomy_mycobank <- taxonomy_mycobank.bak %>% mycobank = `MycoBank #`, fullname = gsub(" +", " ", `Taxon name`), current = clean_name, - ref = paste0(Authors, ", ", `Year of effective publication`), + # without a year, 'Smith, NA' would otherwise become 'Smith et al.' in get_author_year() + ref = if_else( + is.na(`Year of effective publication`), + as.character(Authors), + paste0(Authors, ", ", `Year of effective publication`) + ), rank = `Rank`, status = `Name status`, mycobank_renamed_to = taxonomy_mycobank.bak$`MycoBank #`[match( @@ -735,8 +907,8 @@ taxonomy_mycobank <- taxonomy_mycobank %>% fullname ) ) %>% - # and make fullname distinct again - arrange(rank, mycobank_renamed_to) %>% + # and make fullname distinct again, preferring accepted names (NA would otherwise be sorted last) + arrange(rank, !is.na(mycobank_renamed_to), mycobank_renamed_to) %>% distinct(fullname, .keep_all = TRUE) %>% arrange(fullname) @@ -751,101 +923,27 @@ taxonomy_mycobank <- taxonomy_mycobank %>% taxonomy_mycobank3 <- taxonomy_mycobank taxonomy_mycobank <- taxonomy_mycobank3 -# MycoBank just pasted their taxonomy together into 1 field, and now some classes are in the division (tax_b) column, it's horrible -# so we decide based on the fullname and rank column per record -# use this to determine how far to go: -any( - taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "class"] %in% - taxonomy_mycobank$tax_e -) # replace tax_e with tax_f, etc -any( - taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "class"] %in% - taxonomy_mycobank$tax_f -) +# MycoBank pasted their taxonomy together into 1 field (split into tax_a, tax_b, ... above), and the +# position of each rank differs per record, since e.g. subphyla and subclasses are not always present. +# Previously we only looked in fixed columns per rank, which left e.g. Mucor without phylum, class, +# order and family. So now for every rank, we take the first tax_* column with a known name of that rank. +tax_cols <- grep("^tax_", colnames(taxonomy_mycobank), value = TRUE) +first_name_of_rank <- function(df, rank_name) { + known <- unique(df$fullname[df$rank == rank_name]) + vals <- lapply(df[tax_cols], function(x) if_else(x %in% known, x, NA_character_)) + coalesce(do.call(coalesce, unname(vals)), "") +} taxonomy_mycobank <- taxonomy_mycobank %>% mutate( kingdom = "Fungi", # we already filtered everything else, and MycoBank 90157 has '-' as current name... - phylum = case_when( - rank == "phylum" ~ fullname, - tax_b %in% - taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "phylum"] ~ - tax_b, - tax_c %in% - taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "phylum"] ~ - tax_c, - TRUE ~ "" - ), - class = case_when( - rank == "class" ~ fullname, - tax_c %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "class"] ~ - tax_c, - tax_d %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "class"] ~ - tax_d, - tax_e %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "class"] ~ - tax_e, - TRUE ~ "" - ), - order = case_when( - rank == "order" ~ fullname, - tax_c %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "order"] ~ - tax_c, - tax_d %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "order"] ~ - tax_d, - tax_e %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "order"] ~ - tax_e, - tax_f %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "order"] ~ - tax_f, - tax_g %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "order"] ~ - tax_g, - TRUE ~ "" - ), - family = case_when( - rank == "family" ~ fullname, - tax_c %in% - taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "family"] ~ - tax_c, - tax_d %in% - taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "family"] ~ - tax_d, - tax_e %in% - taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "family"] ~ - tax_e, - tax_f %in% - taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "family"] ~ - tax_f, - tax_g %in% - taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "family"] ~ - tax_g, - tax_h %in% - taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "family"] ~ - tax_h, - tax_i %in% - taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "family"] ~ - tax_i, - TRUE ~ "" - ), + phylum = if_else(rank == "phylum", fullname, first_name_of_rank(taxonomy_mycobank, "phylum")), + class = if_else(rank == "class", fullname, first_name_of_rank(taxonomy_mycobank, "class")), + order = if_else(rank == "order", fullname, first_name_of_rank(taxonomy_mycobank, "order")), + family = if_else(rank == "family", fullname, first_name_of_rank(taxonomy_mycobank, "family")), genus = case_when( rank == "genus" ~ fullname, - tax_b %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "genus"] ~ - tax_b, - tax_c %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "genus"] ~ - tax_c, - tax_d %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "genus"] ~ - tax_d, - tax_e %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "genus"] ~ - tax_e, - tax_f %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "genus"] ~ - tax_f, - tax_g %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "genus"] ~ - tax_g, - tax_h %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "genus"] ~ - tax_h, - tax_i %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "genus"] ~ - tax_i, - tax_j %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "genus"] ~ - tax_j, - tax_k %in% taxonomy_mycobank$fullname[taxonomy_mycobank$rank == "genus"] ~ - tax_k, + # the genus of a (sub)species is always its first word, never rely on the classification for this + rank %in% c("species", "subspecies") ~ gsub(" .*", "", fullname), TRUE ~ "" ), species = case_when( @@ -861,17 +959,31 @@ taxonomy_mycobank <- taxonomy_mycobank %>% TRUE ~ "" ) ) +# these should be very few now: +taxonomy_mycobank %>% + filter(rank %in% c("genus", "species"), family == "") %>% + count(rank) -# use this to get all the genera with updated names from MO_RELEVANT_GENERA: -genera_overview <- AMR::microorganisms %>% - filter(genus %in% AMR:::MO_RELEVANT_GENERA) %>% - pull(fullname) %>% - mo_current() %>% # renamed to's will be included below - mo_genus() %>% - unique() %>% - sort() +# (sub)species without a family in MycoBank get the most common family of their genus +taxonomy_mycobank <- taxonomy_mycobank %>% + group_by(genus) %>% + mutate(across( + c(phylum, class, order, family), + function(x) { + if (genus[1] == "" || all(x != "") || all(x == "")) { + x + } else { + if_else(x == "", names(sort(table(x[x != ""]), decreasing = TRUE))[1], x) + } + } + )) %>% + ungroup() -# keep only the relevant ones +# keep only the relevant genera (see 'Clinically relevant genera' above), but not the fungal homonyms +# of relevant non-fungal genera, such as Necator (otherwise the hookworm would become a fungus) +genera_overview <- relevant_genera[ + !relevant_genera %in% names(genus_domain_override)[genus_domain_override != "Fungi"] +] include <- taxonomy_mycobank %>% filter( genus %in% genera_overview | !rank %in% c("genus", "species", "subspecies") @@ -881,9 +993,18 @@ missing_renamed_to <- include$mycobank_renamed_to[which( !is.na(include$mycobank_renamed_to) & !include$mycobank_renamed_to %in% include$mycobank )] +renamed_to_records <- taxonomy_mycobank %>% + filter(mycobank %in% missing_renamed_to) taxonomy_mycobank <- include %>% - bind_rows(taxonomy_mycobank %>% filter(mycobank %in% missing_renamed_to)) %>% + bind_rows( + renamed_to_records, + # and the genus records of these current names, e.g. a new genus of a renamed relevant species + taxonomy_mycobank %>% + filter(rank == "genus", fullname %in% renamed_to_records$genus, !mycobank %in% include$mycobank) + ) %>% + distinct(mycobank, .keep_all = TRUE) %>% arrange(fullname) +rm(renamed_to_records) # clean up authors and add last columns taxonomy_mycobank <- taxonomy_mycobank %>% @@ -975,9 +1096,9 @@ taxonomy_gbif <- taxonomy_gbif0 %>% c("Archaea", "Bacteria", "Protozoa", na.omit(unique(taxonomy_lpsn$kingdom))) | # include all of these fungal orders order %in% include_fungal_orders | - # and all of these important genera (see "data-raw/_pre_commit_checks.R") + # and all relevant genera (see 'Clinically relevant genera' above) # (they also contain bacteria and protozoa, but these will get higher prevalence scores later on) - genus %in% AMR:::MO_RELEVANT_GENERA + genus %in% relevant_genera ) %>% mutate( # set the right domain for the prokaryotes based on LPSN @@ -987,11 +1108,17 @@ taxonomy_gbif <- taxonomy_gbif0 %>% TRUE ~ kingdom ) ) +# add all synonyms of the included records (most synonyms in GBIF lack their higher taxonomy, so the +# filter above does not catch them) taxonomy_gbif <- taxonomy_gbif %>% - bind_rows(taxonomy_gbif0 %>% - filter(acceptedNameUsageID %in% taxonomy_gbif$taxonID, - !taxonomicStatus %in% c("doubtful", "misapplied", "ambiguous synonym")) %>% - mutate(status = "synonym")) + bind_rows( + taxonomy_gbif0 %>% + filter( + acceptedNameUsageID %in% taxonomy_gbif$taxonID, + !taxonID %in% taxonomy_gbif$taxonID, + !taxonomicStatus %in% c("doubtful", "misapplied", "ambiguous synonym") + ) + ) taxonomy_gbif <- taxonomy_gbif %>% select( @@ -1066,10 +1193,25 @@ accepted <- taxonomy_gbif %>% acc_phylum = phylum, acc_class = class, acc_order = order, - acc_family = family + acc_family = family, + acc_genus = genus ) %>% distinct(gbif, .keep_all = TRUE) +# the higher taxonomy of accepted genera, to be used for synonyms of species in another genus than +# their accepted name (e.g. 'Ascaris cati' is a synonym of 'Toxocara cati', but its family must be the one +# of Ascaris, not of Toxocara) +accepted_genera <- taxonomy_gbif %>% + filter(rank == "genus", status %like% "accepted", !is.na(genus), !is.na(domain)) %>% + distinct(domain, genus, .keep_all = TRUE) %>% + select( + domain, genus, + gen_phylum = phylum, + gen_class = class, + gen_order = order, + gen_family = family + ) + taxonomy_gbif <- taxonomy_gbif %>% left_join(accepted, by = c("gbif_renamed_to" = "gbif")) %>% mutate( @@ -1094,15 +1236,27 @@ taxonomy_gbif <- taxonomy_gbif %>% subspecies = if_else(is.na(subspecies) & rank == "subspecies", sapply(name_parts, `[`, 3L), subspecies), - # Step 2: fill higher taxonomy from accepted name, only for ranks ABOVE the synonym's rank + # Step 2: fill domain and kingdom from the accepted name, only for ranks ABOVE the synonym's rank domain = if_else(is.na(domain) & !is.na(acc_domain), acc_domain, domain), kingdom = if_else(is.na(kingdom) & !is.na(acc_kingdom) & rank_num > 1L, acc_kingdom, kingdom), - phylum = if_else(is.na(phylum) & !is.na(acc_phylum) & rank_num > 2L, acc_phylum, phylum), - class = if_else(is.na(class) & !is.na(acc_class) & rank_num > 3L, acc_class, class), - order = if_else(is.na(order) & !is.na(acc_order) & rank_num > 4L, acc_order, order), - family = if_else(is.na(family) & !is.na(acc_family) & rank_num > 5L, acc_family, family) + # Step 3: phylum to family may only come from the accepted name if that is in the same genus + # (or if the synonym is a genus or higher), otherwise from the synonym's own genus (see Step 4) + same_lineage = rank_num <= 6L | (!is.na(acc_genus) & genus == acc_genus), + phylum = if_else(is.na(phylum) & !is.na(acc_phylum) & rank_num > 2L & same_lineage, acc_phylum, phylum), + class = if_else(is.na(class) & !is.na(acc_class) & rank_num > 3L & same_lineage, acc_class, class), + order = if_else(is.na(order) & !is.na(acc_order) & rank_num > 4L & same_lineage, acc_order, order), + family = if_else(is.na(family) & !is.na(acc_family) & rank_num > 5L & same_lineage, acc_family, family) ) %>% - select(-starts_with("acc_"), -name_parts, -rank_num) + # Step 4: species in another genus than their accepted name get the taxonomy of their own genus + left_join(accepted_genera, by = c("domain", "genus")) %>% + mutate( + phylum = if_else(is.na(phylum) & rank_num > 6L, gen_phylum, phylum), + class = if_else(is.na(class) & rank_num > 6L, gen_class, class), + order = if_else(is.na(order) & rank_num > 6L, gen_order, order), + family = if_else(is.na(family) & rank_num > 6L, gen_family, family) + ) %>% + select(-starts_with("acc_"), -starts_with("gen_"), -name_parts, -rank_num, -same_lineage) +rm(accepted, accepted_genera) taxonomy_gbif <- taxonomy_gbif %>% @@ -1127,6 +1281,18 @@ taxonomy_gbif <- readRDS("data-raw/taxonomy_gbif.rds") # Add full names ---------------------------------------------------------------------------------- +# names must be ASCII before they are made unique, otherwise e.g. 'Fusarium aloës' and 'Fusarium aloes' +# both end up as 'Fusarium aloes' in the final data set (which happened to 12 names in 2026) +ascii_names <- function(x) { + gsub("[\"'`]", "", stringi::stri_trans_general(x, "Latin-ASCII")) +} +taxonomy_gbif <- taxonomy_gbif %>% + mutate(across(domain:subspecies, ascii_names)) +taxonomy_lpsn <- taxonomy_lpsn %>% + mutate(across(domain:subspecies, ascii_names)) +taxonomy_mycobank <- taxonomy_mycobank %>% + mutate(across(domain:subspecies, ascii_names)) + taxonomy_gbif <- taxonomy_gbif %>% # clean NAs and add fullname mutate( @@ -1207,13 +1373,12 @@ taxonomy <- taxonomy_lpsn %>% select(fullname, everything()) # get missing entries from existing microorganisms data set -taxonomy.old <- AMR::microorganisms %>% +# (only names that are completely absent now, so that a taxon that moved to another domain, such as +# Trichophyton in 2026, does not come back in its old domain) +taxonomy.old <- microorganisms_old %>% select(any_of(colnames(taxonomy))) %>% - # TODO ONLY IN 2026! - mutate(domain = kingdom) %>% - # END OF TODO filter( - !paste(domain, fullname) %in% paste(taxonomy$domain, taxonomy$fullname), + !fullname %in% taxonomy$fullname, # these will be added later: tolower(source) != "manually added" ) @@ -1243,18 +1408,25 @@ taxonomy %>% count(rank, sort = TRUE) # Resolve genera that appear in multiple domains (e.g. Giardia in Protozoa # and Animalia, or Trichophyton in Fungi and as a cyanobacterium in GBIF). -# The winning lineage per genus is decided in this order: -# 1. Authoritative sources (LPSN, MycoBank) always beat GBIF and other sources, +# The winning domain per genus is decided in this order: +# 1. Our own overrides in `genus_domain_override` (see 'Clinically relevant genera' above). +# 2. Authoritative sources (LPSN, MycoBank) always beat GBIF and other sources, # so GBIF can never overrule LPSN or MycoBank. -# 2. Accepted beats non-accepted, based on the genus-rank record of that -# source/domain (or its species if the genus record is missing). So a -# synonymic bacterial genus in LPSN loses from an accepted fungal genus in +# 3. Within LPSN and MycoBank: accepted beats non-accepted, based on the genus-rank +# record of that source/domain (or its species if the genus record is missing). +# So a synonymic bacterial genus in LPSN loses from an accepted fungal genus in # MycoBank with the same name (e.g. Pirella), and vice versa. -# 3. Source: LPSN > MycoBank > GBIF > other. -# 4. Genus-rank records over species/subspecies records. -# 5. Domain: Bacteria > Fungi > Protozoa > Archaea > Chromista > Animalia > +# 4. Source: LPSN > MycoBank > GBIF > other. +# 5. Genus-rank records over species/subspecies records. +# 6. Domain: Bacteria > Fungi > Protozoa > Archaea > Chromista > Animalia > # Plantae, which only breaks ties within the same source (e.g. Giardia within GBIF). -# The winning lineage then overwrites domain-to-family for the entire genus. +# The lineage of the winning domain then overwrites domain-to-family for the entire genus. +# Records of the losing domain are: +# - moved to the winning domain if they are protists that were placed in another kingdom (GBIF places +# e.g. some Trypanosoma, Giardia and Amoeba species in Animalia), or if they have the same phylum or +# class as the winning lineage (e.g. microsporidia placed in Protozoa instead of Fungi); +# - removed in all other cases, as they are other organisms with the same genus name (e.g. the +# Graphium butterflies vs the Graphium moulds, or the Pirella bacteria vs the Pirella fungi). taxonomy %>% count(domain, sort = TRUE) domain_priority <- c( "Bacteria" = 1L, @@ -1265,11 +1437,11 @@ domain_priority <- c( "Animalia" = 6L, "Plantae" = 7L ) -source_priority <- c("LPSN" = 1L, "MycoBank" = 2L, "GBIF" = 3L, "manually added" = 4L) +first_nonempty <- function(x) { + coalesce(x[!x %in% c("", NA)][1], "") +} -# find the best domain per genus, from all records of that genus (so that e.g. a -# MycoBank species also counts when the genus-rank record only came from GBIF) -best_domain <- taxonomy %>% +genus_candidates <- taxonomy %>% filter( genus != "", rank %in% c("genus", "species", "subspecies"), @@ -1283,51 +1455,94 @@ best_domain <- taxonomy %>% mutate( # status of this genus in this source/domain: from the genus-rank record if # available, otherwise accepted if any of its species/subspecies is accepted - genus_status = if (any(rank == "genus")) { - status[rank == "genus"][1] - } else if (any(status == "accepted")) { - "accepted" - } else { - status[1] - } + genus_status = coalesce( + first(status[rank == "genus"]), + if_else(any(status == "accepted"), "accepted", first(status)) + ) ) %>% ungroup() %>% mutate( + override = unname(genus_domain_override[genus]), + oprio = case_when( + is.na(override) ~ 1L, + domain == override ~ 0L, + TRUE ~ 2L + ), tprio = if_else(source %in% c("LPSN", "MycoBank"), 1L, 2L), - stprio = if_else(genus_status %in% "accepted", 1L, 2L), - sprio = coalesce(unname(source_priority[source]), 5L), + stprio = if_else(tprio == 2L | genus_status %in% "accepted", 1L, 2L), + sprio = source_prio(source), rprio = if_else(rank == "genus", 1L, 2L), dprio = unname(domain_priority[domain]) ) %>% - arrange(genus, tprio, stprio, sprio, rprio, dprio) %>% - distinct(genus, .keep_all = TRUE) %>% - select(genus, best_domain = domain, best_kingdom = kingdom, - best_phylum = phylum, best_class = class, - best_order = order, best_family = family) + arrange(genus, oprio, tprio, stprio, sprio, rprio, dprio) + +best_domain <- genus_candidates %>% + group_by(genus) %>% + mutate(best_domain = first(domain)) %>% + # take every rank from the most reliable record that has a value for it, so that an empty family + # in the winning record (e.g. Mucor in MycoBank) cannot erase the family from another source + filter(domain == best_domain) %>% + summarise( + best_domain = first(best_domain), + best_kingdom = first_nonempty(kingdom), + best_phylum = first_nonempty(phylum), + best_class = first_nonempty(class), + best_order = first_nonempty(order), + best_family = first_nonempty(family), + .groups = "drop" + ) + +# REVIEW: genus homonyms of clinically relevant genera, check the winning domain and extend +# `genus_domain_override` at the start of this script if it is wrong +genus_candidates %>% + filter(genus %in% relevant_genera) %>% + group_by(genus) %>% + filter(n_distinct(domain) > 1) %>% + group_by(genus, domain) %>% + summarise( + n = n(), + sources = paste(sort(unique(source)), collapse = "+"), + example = first(fullname[rank == "species"]), + .groups = "drop" + ) %>% + left_join(best_domain %>% select(genus, best_domain), by = "genus") %>% + mutate(in_override_list = genus %in% names(genus_domain_override)) %>% + review("Homonyms of relevant genera (check `best_domain`)") taxonomy <- taxonomy %>% left_join(best_domain, by = "genus") %>% - # LPSN and MycoBank records of a losing homonym are different organisms under - # another nomenclatural code, so remove them instead of moving them to the - # winning domain (e.g. the bacterial Pirella species if the fungal Pirella wins) - filter( - !(genus != "" & !is.na(best_domain) & source == "LPSN" & - !best_domain %in% c("Bacteria", "Archaea")), - !(genus != "" & !is.na(best_domain) & source == "MycoBank" & - best_domain != "Fungi") - ) %>% mutate( + .resolve = case_when( + genus == "" | is.na(best_domain) | !domain %in% names(domain_priority) ~ "keep", + domain == best_domain ~ "keep", + # LPSN and MycoBank records of a losing homonym are other organisms under another + # nomenclatural code (e.g. the bacterial Pirella species if the fungal Pirella wins) + source %in% c("LPSN", "MycoBank") ~ "remove", + # protists placed in another kingdom + best_domain %in% c("Protozoa", "Chromista") & domain %in% c("Protozoa", "Chromista", "Animalia") ~ "move", + # same phylum or class, so the same organisms in another kingdom (e.g. microsporidia as Protozoa + # instead of Fungi, which share the class Microsporea) + (phylum != "" & phylum == best_phylum) | (class != "" & class == best_class) ~ "move", + TRUE ~ "remove" + ) + ) +review( + taxonomy %>% filter(.resolve == "remove") %>% count(genus, domain, best_domain, source, sort = TRUE), + "Records removed as homonyms of a genus in another domain" +) +taxonomy <- taxonomy %>% + filter(.resolve != "remove") %>% + mutate( + is_genus_rank = genus != "" & !is.na(best_domain) & rank %in% c("genus", "species", "subspecies"), domain = if_else(genus != "" & !is.na(best_domain), best_domain, domain), - kingdom = if_else(genus != "" & !is.na(best_kingdom), best_kingdom, kingdom), - phylum = if_else(genus != "" & !is.na(best_phylum) & rank %in% - c("genus", "species", "subspecies"), best_phylum, phylum), - class = if_else(genus != "" & !is.na(best_class) & rank %in% - c("genus", "species", "subspecies"), best_class, class), - order = if_else(genus != "" & !is.na(best_order) & rank %in% - c("genus", "species", "subspecies"), best_order, order), - family = if_else(genus != "" & !is.na(best_family) & rank %in% - c("genus", "species", "subspecies"), best_family, family)) %>% - select(-starts_with("best_")) + kingdom = if_else(genus != "" & !is.na(best_kingdom) & best_kingdom != "", best_kingdom, kingdom), + phylum = if_else(is_genus_rank & best_phylum != "", best_phylum, phylum), + class = if_else(is_genus_rank & best_class != "", best_class, class), + order = if_else(is_genus_rank & best_order != "", best_order, order), + family = if_else(is_genus_rank & best_family != "", best_family, family) + ) %>% + select(-starts_with("best_"), -is_genus_rank, -.resolve) +rm(genus_candidates, best_domain) # integrity tests: no LPSN record may have left the prokaryotes and no MycoBank # record may have left the fungi (see issue #309, Trichophyton as cyanobacterium). @@ -1340,7 +1555,10 @@ taxonomy %>% count(domain, genus, sort = TRUE) # and these known cross-domain homonyms must end up in the right domain taxonomy %>% - filter(rank == "genus", genus %in% c("Trichophyton", "Pirella", "Tubercularia")) %>% + filter( + rank == "genus", + genus %in% c("Trichophyton", "Pirella", "Tubercularia", "Necator", "Capillaria", "Graphium", "Trypanosoma") + ) %>% select(fullname, domain, kingdom, phylum, family, source) # check if some genera within kingdoms have multiple families / orders, etc: @@ -1348,7 +1566,8 @@ taxonomy %>% filter(genus != "") %>% group_by(kingdom, genus) %>% filter(n_distinct(family) > 1) %>% - View() + ungroup() %>% + review("Genera with multiple families") # then fix (we're still arranged by domain, source here): taxonomy <- taxonomy %>% group_by(kingdom, genus) %>% @@ -1404,7 +1623,7 @@ taxonomy <- taxonomy %>% TRUE ~ paste(genus, species, subspecies) # already trimmed 7 lines up ))) %>% # keep the record of the most authoritative source: LPSN > MycoBank > GBIF > other - mutate(sprio = coalesce(unname(source_priority[source]), 5L)) %>% + mutate(sprio = source_prio(source)) %>% arrange(fullname, sprio) %>% distinct(fullname, .keep_all = TRUE) %>% select(-sprio) @@ -1429,90 +1648,44 @@ current_gbif <- taxonomy_gbif.bak %>% kingdom %in% na.omit(taxonomy_lpsn$kingdom[taxonomy_lpsn$domain == "Bacteria"]) ~ "Bacteria", kingdom %in% na.omit(taxonomy_lpsn$kingdom[taxonomy_lpsn$domain == "Archaea"]) ~ "Archaea", TRUE ~ kingdom - ) - ) - -rank_levels <- c("domain", "kingdom", "phylum", "class", "order", "family", "genus") - -taxonomy_all_missing <- map_dfr(rank_levels, function(rank_name) { - candidates <- taxonomy %>% - filter(.data[[rank_name]] != "") %>% - distinct(across(all_of(unique(c("domain", rank_levels[seq_len(which(rank_levels == rank_name))]))))) %>% - mutate(fullname = .data[[rank_name]], rank = rank_name) %>% - anti_join( - taxonomy %>% filter(rank == rank_name), - by = unique(c("domain", setNames(rank_name, rank_name), "rank")) - ) - - if (nrow(candidates) == 0) return(tibble()) - - message("Adding missing: ", rank_name, "... n = ", nrow(candidates)) - - # enrich from GBIF - gbif_lookup <- current_gbif %>% - filter(taxonRank == rank_name) %>% - select(all_of(unique(c("domain", rank_name))), - ref = scientificNameAuthorship, - gbif = taxonID, - gbif_parent = parentNameUsageID) %>% - distinct() - - candidates %>% - left_join(gbif_lookup, by = unique(c("domain", rank_name))) %>% - mutate( - source = if_else(!is.na(gbif), "GBIF", "manually added"), - status = if_else(!is.na(gbif), "accepted", "unknown") - ) -}) - -# species implied by subspecies but missing as a species-rank row -missing_species <- taxonomy %>% - filter(species != "") %>% - distinct(domain, genus, species, .keep_all = TRUE) %>% - select(domain:species) %>% - mutate(fullname = paste(genus, species), rank = "species") %>% - anti_join( - taxonomy %>% filter(rank == "species"), - by = c("domain", "genus", "species", "rank") + ), + # the own rank of a record is not always filled in, so take it from the name (as for `taxonomy_gbif`) + canonical = case_when( + is.na(scientificNameAuthorship) | scientificNameAuthorship == "" ~ scientificName, + endsWith(scientificName, scientificNameAuthorship) ~ + trimws(substr(scientificName, 1L, nchar(scientificName) - nchar(scientificNameAuthorship))), + TRUE ~ scientificName + ), + genus = if_else(taxonRank == "genus", canonical, genus), + family = if_else(taxonRank == "family", canonical, family), + order = if_else(taxonRank == "order", canonical, order), + class = if_else(taxonRank == "class", canonical, class), + phylum = if_else(taxonRank == "phylum", canonical, phylum) ) %>% - left_join( - current_gbif %>% - filter(taxonRank == "species") %>% - select(domain, genus, species = specificEpithet, - ref = scientificNameAuthorship, - gbif = taxonID, gbif_parent = parentNameUsageID) %>% - distinct(), - by = c("domain", "genus", "species") - ) %>% - mutate( - source = if_else(!is.na(gbif), "GBIF", "manually added"), - status = if_else(!is.na(gbif), "accepted", "unknown") - ) + select(-canonical) -taxonomy <- taxonomy %>% - bind_rows(taxonomy_all_missing, missing_species) %>% - mutate( - domain = if_else(is.na(domain) | domain == "", kingdom, domain), - across(kingdom:subspecies, \(x) if_else(is.na(x), "", x)) - ) +taxonomy <- add_missing_parents(taxonomy, current_gbif) # TODO perhaps only for 2026: remove old kingdoms -taxonomy <- taxonomy %>% +taxonomy <- taxonomy %>% filter(!(rank == "kingdom" & fullname %in% c("Bacteria", "Archaea"))) # Deduplicate: where the same fullname appears at multiple ranks within a # domain (e.g. Nitrospira as genus and class), keep the lowest rank as-is # and disambiguate higher ranks by appending {rank}. -rank_priority <- c("subspecies" = 1L, "species" = 2L, "genus" = 3L, - "family" = 4L, "order" = 5L, "class" = 6L, - "phylum" = 7L, "kingdom" = 8L, "domain" = 9L) -source_priority <- c("LPSN" = 1L, "MycoBank" = 2L, "GBIF" = 3L, "manually added" = 4L) +rank_priority <- c( + "subspecies" = 1L, "species" = 2L, "genus" = 3L, + "family" = 4L, "order" = 5L, "class" = 6L, + "phylum" = 7L, "kingdom" = 8L, "domain" = 9L +) taxonomy <- taxonomy %>% - mutate(rank_index = rank_priority[rank], - source_index = source_priority[source]) %>% + mutate( + rank_index = rank_priority[rank], + source_index = source_prio(source) + ) %>% arrange(domain, fullname, rank_index, source_index) %>% - distinct(domain, fullname, rank_index, .keep_all = TRUE) %>% + distinct(domain, fullname, rank_index, .keep_all = TRUE) %>% group_by(domain, fullname) %>% mutate(fullname = if_else( row_number() > 1, @@ -1534,21 +1707,21 @@ saveRDS(taxonomy, "data-raw/taxonomy1.rds") # Get previously manually added entries ----------------------------------------------------------- -manually_added <- AMR::microorganisms %>% +# (note: until 2026, also records that were only inferred by this script were labelled 'manually added', +# so these are not all curated entries - they follow the relevance filters in 'Remove unwanted taxonomic entries') +manually_added <- microorganisms_old %>% filter( tolower(source) == "manually added", !paste(kingdom, fullname) %in% paste(taxonomy$kingdom, taxonomy$fullname), !rank %in% c("domain", "kingdom", "phylum", "class", "order", "family") ) %>% - select(fullname:subspecies, ref, source, rank) + select(fullname:subspecies, ref, source, rank, status) # Build a lookup for each child->parent rank pair, preferring LPSN > MycoBank > any -source_priority <- c("LPSN" = 1L, "MycoBank" = 2L, "GBIF" = 3L, "manually added" = 4L) - fill_from_parent <- function(data, taxonomy, child_rank, parent_rank) { lookup <- taxonomy %>% filter(.data[[child_rank]] != "", .data[[parent_rank]] != "") %>% - mutate(sprio = source_priority[source]) %>% + mutate(sprio = source_prio(source)) %>% arrange(.data[[child_rank]], sprio) %>% distinct(.data[[child_rank]], .keep_all = TRUE) %>% select(all_of(c(child_rank, parent_rank))) @@ -1572,7 +1745,8 @@ for (i in seq_len(nrow(rank_pairs))) { manually_added <- manually_added %>% mutate( - status = "unknown", + # keep a curated 'accepted' status (such as of Plasmodium), all others are unknown + status = if_else(status %in% "accepted", status, "unknown"), rank = if_else(fullname %like% "unknown", "(unknown rank)", rank) ) %>% filter(!fullname %in% taxonomy$fullname) @@ -1612,7 +1786,7 @@ lpsn_genera <- taxonomy %>% pull(genus) gbif_bacteria <- which( taxonomy$domain == "Bacteria" & - taxonomy$source %in% c("GBIF", "manually added") & + taxonomy$source %in% c("GBIF", "manually added", "inferred") & (taxonomy$rank %in% c("phylum", "class", "order", "family", "genus") | (taxonomy$rank == "species" & taxonomy$genus %in% lpsn_genera)) ) @@ -1622,9 +1796,12 @@ pb <- progress_bar$new( total = length(gbif_bacteria), format = "[:bar] :current/:total :eta" ) -for (record in gbif_bacteria) { +for (i in seq_along(gbif_bacteria)) { pb$tick() - Sys.sleep(0.1) + record <- gbif_bacteria[i] + if (i %% 500 == 0) { + save_lpsn_cache() + } lpsn <- get_lpsn_and_author( rank = taxonomy$rank[record], name = taxonomy$fullname[record] @@ -1639,6 +1816,7 @@ for (record in gbif_bacteria) { taxonomy$status[record] <- unname(lpsn["status"]) } } +save_lpsn_cache() warnings() message(added, " GBIF records altered to latest LPSN") @@ -1703,6 +1881,11 @@ for (d in unique(taxonomy$domain[taxonomy$domain != "(unknown domain)"])) { } lookup <- lookup %>% + # an empty parent in the authoritative source must never erase a known parent from GBIF + # (e.g. MycoBank has no family for Mucor, GBIF has Mucoraceae) + filter(!.data[[parent_col]] %in% c("", NA)) %>% + # prefer the record of the child itself, e.g. the family record for the order of a family + arrange(rank != child_col) %>% distinct(.data[[child_col]], .keep_all = TRUE) %>% select(all_of(c(child_col, parent_col))) @@ -1752,8 +1935,23 @@ saveRDS(taxonomy, "data-raw/taxonomy1c.rds") # Add parent identifiers -------------------------------------------------------------------------- taxonomy$source[taxonomy$source == "Manually added"] <- "manually added" +# Names that are not validly published (according to LPSN) or could not be checked are removed, unless +# they are clinically relevant. E.g., Tropheryma whipplei is 'not validly published' under the ICNP, and +# was therefore removed from this data set in 2026. +# For prokaryotes, only the human pathogens of Bartlett et al. (and their genera) are protected, otherwise all +# non-validly published species of e.g. Streptococcus would return. +protected <- (taxonomy$domain %in% c("Bacteria", "Archaea") & + (paste(taxonomy$genus, taxonomy$species) %in% paste(pathogens$genus, pathogens$species) | + (taxonomy$rank == "genus" & taxonomy$genus %in% pathogens$genus))) | + (!taxonomy$domain %in% c("Bacteria", "Archaea") & taxonomy$genus %in% relevant_genera) +review( + taxonomy %>% filter(!status %in% c("accepted", "synonym"), protected, !source %in% c("manually added", "inferred")), + "Not validly published or unknown status, but kept as clinically relevant" +) taxonomy <- taxonomy %>% - filter(status %in% c("accepted", "synonym") | source == "manually added") + mutate(status = if_else(!status %in% c("accepted", "synonym") & protected & source != "inferred", "accepted", status)) %>% + filter(status %in% c("accepted", "synonym") | source %in% c("manually added", "inferred")) +rm(protected) # Add domain rows taxonomy <- taxonomy %>% @@ -1778,7 +1976,7 @@ taxonomy$kingdom[taxonomy$kingdom == taxonomy$domain & taxonomy$rank == "domain"] <- "" taxonomy <- taxonomy %>% - arrange(fullname, lpsn, mycobank, desc(gbif)) %>% + arrange(fullname, source_prio(source), lpsn, mycobank, desc(gbif)) %>% distinct(fullname, .keep_all = TRUE) %>% arrange(fullname) @@ -1835,8 +2033,7 @@ taxonomy$rank[is.na(taxonomy$rank)] <- "domain" # Add prevalence ---------------------------------------------------------------------------------- # this part is required here, because it's needed for filtering on 'relevant' species to keep later on - -pathogens <- read_excel(file_bartlett, sheet = "Tab 6 Full List") +# (`pathogens` was read at the start of this script) # get all established, both old and current taxonomic names established <- pathogens %>% @@ -1896,42 +2093,61 @@ nonbacterial_genera <- nonbacterial_genera[ # update prevalence based on taxonomy (following the recent and thorough work of Bartlett et al., 2022) # see https://doi.org/10.1099/mic.0.001269 -taxonomy <- taxonomy %>% - mutate( - prevalence = case_when( - # genera of the pathogens mentioned in the World Health Organization's (WHO) Priority Pathogen List - genus %in% MO_WHO_PRIORITY_GENERA ~ 1.0, - # 'established' means 'have infected at least three persons in three or more references' - paste(genus, species) %in% - established & - rank %in% c("species", "subspecies") ~ - 1.15, - # other genera in the 'established' group - genus %in% established_genera & rank == "genus" ~ 1.15, - - # 'putative' means 'fewer than three known cases' - paste(genus, species) %in% - putative & - rank %in% c("species", "subspecies") ~ - 1.25, - - # other genera in the 'putative' group - genus %in% putative_genera & rank == "genus" ~ 1.25, - # species and subspecies in 'established' and 'putative' groups - genus %in% - c(established_genera, putative_genera) & - rank %in% c("species", "subspecies") ~ - 1.5, - - # non-bacterial genera/species/subspecies of clinical relevance - genus %in% nonbacterial_genera & - domain != "Bacteria" & - rank %in% c("genus", "species", "subspecies") ~ - 1.25, - # all others - TRUE ~ 2.0 +# this is a function, since it must be run again after the higher taxonomy has been harmonised below +compute_prevalence <- function(taxonomy) { + taxonomy <- taxonomy %>% + mutate( + who_genus = genus %in% MO_WHO_PRIORITY_GENERA, + prevalence = case_when( + # 'established' means 'have infected at least three persons in three or more references' + paste(genus, species) %in% established & rank %in% c("species", "subspecies") ~ 1.15, + # other genera in the 'established' group + genus %in% established_genera & rank == "genus" ~ 1.15, + # 'putative' means 'fewer than three known cases' + paste(genus, species) %in% putative & rank %in% c("species", "subspecies") ~ 1.25, + # other genera in the 'putative' group + genus %in% putative_genera & rank == "genus" ~ 1.25, + # species and subspecies in 'established' and 'putative' groups + genus %in% c(established_genera, putative_genera) & rank %in% c("species", "subspecies") ~ 1.5, + # non-bacterial genera/species/subspecies of clinical relevance + genus %in% nonbacterial_genera & domain != "Bacteria" & rank %in% c("genus", "species", "subspecies") ~ 1.25, + # all others + TRUE ~ 2.0 + ) ) - ) + # a current name is as relevant as its synonyms, so that e.g. Candidozyma auris (renamed from + # Candida auris in 2024) gets the prevalence of Candida auris, without having to update any list. + # This uses the prevalence without the WHO rule below, since that rule applies to whole genera: + # otherwise every species that ever was in e.g. Pseudomonas would become 1.0. + for (id in c("lpsn", "mycobank", "gbif")) { + renamed_col <- paste0(id, "_renamed_to") + if (!all(c(id, renamed_col) %in% colnames(taxonomy))) { + next + } + via_synonyms <- taxonomy %>% + filter(!is.na(.data[[renamed_col]])) %>% + group_by(target = .data[[renamed_col]]) %>% + summarise(syn_prevalence = min(prevalence), .groups = "drop") + taxonomy <- taxonomy %>% + left_join(via_synonyms, by = setNames("target", id), na_matches = "never") %>% + mutate(prevalence = pmin(prevalence, syn_prevalence, na.rm = TRUE)) %>% + select(-syn_prevalence) + } + # genera of the pathogens mentioned in the World Health Organization's (WHO) Priority Pathogen List + taxonomy <- taxonomy %>% + mutate(prevalence = if_else(who_genus, 1.0, prevalence)) %>% + select(-who_genus) + # and a genus is as relevant as its most relevant accepted species + genus_prevalence <- taxonomy %>% + filter(rank == "species", status == "accepted", genus != "") %>% + group_by(domain, genus) %>% + summarise(gen_prevalence = min(prevalence), .groups = "drop") + taxonomy %>% + left_join(genus_prevalence, by = c("domain", "genus")) %>% + mutate(prevalence = if_else(rank == "genus", pmin(prevalence, gen_prevalence, na.rm = TRUE), prevalence)) %>% + select(-gen_prevalence) +} +taxonomy <- compute_prevalence(taxonomy) table(taxonomy$prevalence, useNA = "always") # (a lot will be removed further below) @@ -1947,8 +2163,9 @@ saveRDS(taxonomy, "data-raw/taxonomy2.rds") part1 <- taxonomy %>% filter( - # keep all we added ourselves: - source == "manually added" | + # keep all unknowns we added ourselves (other 'manually added' records follow the + # rules below, since until 2026 also records that were only inferred were labelled 'manually added') + fullname %like% "unknown" | # keep all bacteria anyway, main focus of our package: domain == "Bacteria" | # and these kingdoms are very small, and also mainly microorganisms: @@ -1957,8 +2174,8 @@ part1 <- taxonomy %>% domain == "Chromista" | # keep everything from family up, 'ghost' entries will be removed later on rank %in% c("domain", "kingdom", "phylum", "class", "order", "family") | - # other relevant genera to keep: - genus %in% AMR:::MO_RELEVANT_GENERA | + # other relevant genera to keep (see 'Clinically relevant genera' above): + genus %in% relevant_genera | # relevant for biotechnology: genus %in% c( @@ -1999,70 +2216,40 @@ part1 <- taxonomy %>% # this domain only contained Curvularia and Hymenolepis, which have coincidental twin names with Fungi filter( domain != "Plantae", + # species groups are added again at the end from the previous data set, and would otherwise get the + # MO code of their genus below (e.g. B_MYCBC for 'Mycobacterium avium-intracellulare complex') + rank != "species group", !(genus %in% c("Aedes", "Anopheles") & rank %in% c("species", "subspecies")) ) -# now get the parents and old names -part2 <- taxonomy %>% - filter( - gbif %in% - c( - part1$gbif_parent[!is.na(part1$gbif_parent)], - part1$gbif_renamed_to[!is.na(part1$gbif_renamed_to)] - ) | - mycobank %in% - c( - part1$mycobank_parent[!is.na(part1$mycobank_parent)], - part1$mycobank_renamed_to[!is.na(part1$mycobank_renamed_to)] - ) | - lpsn %in% - c( - part1$lpsn_parent[!is.na(part1$lpsn_parent)], - part1$lpsn_renamed_to[!is.na(part1$lpsn_renamed_to)] - ) - ) -parts <- bind_rows(part1, part2) - -part3 <- taxonomy %>% - filter( - gbif %in% - c( - parts$gbif_parent[!is.na(parts$gbif_parent)], - parts$gbif_renamed_to[!is.na(parts$gbif_renamed_to)] - ) | - mycobank %in% - c( - parts$mycobank_parent[!is.na(parts$mycobank_parent)], - parts$mycobank_renamed_to[!is.na(parts$mycobank_renamed_to)] - ) | - lpsn %in% - c( - parts$lpsn_parent[!is.na(parts$lpsn_parent)], - parts$lpsn_renamed_to[!is.na(parts$lpsn_renamed_to)] - ) - ) -parts <- bind_rows(part1, part2, part3) - -part4 <- taxonomy %>% - filter( - gbif %in% - c( - parts$gbif_parent[!is.na(parts$gbif_parent)], - parts$gbif_renamed_to[!is.na(parts$gbif_renamed_to)] - ) | - mycobank %in% - c( - parts$mycobank_parent[!is.na(parts$mycobank_parent)], - parts$mycobank_renamed_to[!is.na(parts$mycobank_renamed_to)] - ) | - lpsn %in% - c( - parts$lpsn_parent[!is.na(parts$lpsn_parent)], - parts$lpsn_renamed_to[!is.na(parts$lpsn_renamed_to)] - ) - ) -parts <- bind_rows(part1, part2, part3, part4) - +# now get the parents, the current names of kept synonyms, and the synonyms of kept (sub)species, +# and repeat until nothing is added anymore (previously this was done a fixed number of 3 times) +parts <- part1 +repeat { + ids_wanted <- function(id) { + na.omit(c(parts[[paste0(id, "_parent")]], parts[[paste0(id, "_renamed_to")]])) + } + kept_species <- parts %>% filter(rank %in% c("species", "subspecies")) + added <- taxonomy %>% + filter( + !fullname %in% parts$fullname, + # parents and current names + gbif %in% ids_wanted("gbif") | + mycobank %in% ids_wanted("mycobank") | + lpsn %in% ids_wanted("lpsn") | + # old names of kept (sub)species, so that as.mo() can still find e.g. old fungal names + (status == "synonym" & rank %in% c("species", "subspecies") & + (gbif_renamed_to %in% na.omit(kept_species$gbif) | + mycobank_renamed_to %in% na.omit(kept_species$mycobank) | + lpsn_renamed_to %in% na.omit(kept_species$lpsn))) + ) + message("Adding ", nrow(added), " parents, current names and synonyms") + if (nrow(added) == 0) { + break + } + parts <- bind_rows(parts, added) +} +rm(added, kept_species) # Keep accepted names that any kept synonym points to accepted_targets <- taxonomy %>% @@ -2134,11 +2321,10 @@ genus_ref <- taxonomy %>% summarise( .ref_domain = coalesce(first(domain[rank == "genus"]), first(domain)), .ref_kingdom = coalesce(first(kingdom[rank == "genus"]), first(kingdom)), - .ref_phylum = coalesce(first(phylum[rank == "genus"]), first(phylum)), - .ref_class = coalesce(first(class[rank == "genus"]), first(class)), - .ref_order = coalesce(first(order[rank == "genus"]), first(order)), - .ref_family = coalesce(first(family[rank == "genus"]), first(family)), - .ref_prevalence = coalesce(first(prevalence[rank == "genus"]), first(prevalence)), + .ref_phylum = consensus_value(phylum, rank, source, "genus"), + .ref_class = consensus_value(class, rank, source, "genus"), + .ref_order = consensus_value(order, rank, source, "genus"), + .ref_family = consensus_value(family, rank, source, "genus"), .groups = "drop" ) @@ -2148,40 +2334,28 @@ taxonomy <- taxonomy %>% across_flag = rank != "genus" & genus != "" & fullname %unlike% "unknown", domain = if_else(across_flag, .ref_domain, domain), kingdom = if_else(across_flag, .ref_kingdom, kingdom), - phylum = if_else(across_flag, .ref_phylum, phylum), - class = if_else(across_flag, .ref_class, class), - order = if_else(across_flag, .ref_order, order), - family = if_else(across_flag, .ref_family, family), - prevalence = if_else(across_flag, .ref_prevalence, prevalence) + # also fill the genus record itself where it has no value + phylum = if_else((across_flag | (rank == "genus" & phylum == "")) & .ref_phylum != "", .ref_phylum, phylum), + class = if_else((across_flag | (rank == "genus" & class == "")) & .ref_class != "", .ref_class, class), + order = if_else((across_flag | (rank == "genus" & order == "")) & .ref_order != "", .ref_order, order), + family = if_else((across_flag | (rank == "genus" & family == "")) & .ref_family != "", .ref_family, family) ) %>% select(-starts_with(".ref_"), -across_flag) %>% arrange(fullname) +# (until 2026, the prevalence of the genus was also copied to all its species here, which made the +# species-level information of Bartlett et al. useless - the prevalence is now recalculated below) # update the taxonomic names based on new genus classification taxonomy <- taxonomy %>% - # first, fix kingdom based on family level (only when family is not empty) + # first, fix the higher taxonomy of families, preferring the family record itself, then the most + # authoritative source (previously the first record in alphabetical order was taken, so that one outdated + # GBIF record could change the phylum of a whole family) group_by(domain, family) %>% mutate( - kingdom = if_else( - family != "", - get_top_lvl(kingdom, rank, source, "phylum", phylum), - kingdom - ), - phylum = if_else( - family != "", - first(phylum[phylum != "" & rank != "phylum"]), - phylum - ), - class = if_else( - family != "", - first(class[class != "" & rank != "class"]), - class - ), - order = if_else( - family != "", - first(order[order != "" & rank != "order"]), - order - ) + kingdom = if_else(family != "", consensus_value(kingdom, rank, source, "family"), kingdom), + phylum = if_else(family != "", consensus_value(phylum, rank, source, "family"), phylum), + class = if_else(family != "", consensus_value(class, rank, source, "family"), class), + order = if_else(family != "", consensus_value(order, rank, source, "family"), order) ) %>% # next, update all taxonomic layers group_by(domain, genus) %>% @@ -2196,6 +2370,9 @@ taxonomy <- taxonomy %>% arrange(fullname) %>% mutate(across(kingdom:family, ~ if_else(is.na(.x), "", .x))) +# recalculate the prevalence, now that all species have the taxonomy of their genus +taxonomy <- compute_prevalence(taxonomy) + # no ghost families, orders, classes, phyla # (but keep the ghost families of bacteria) taxonomy <- taxonomy %>% @@ -2235,6 +2412,22 @@ taxonomy <- taxonomy %>% # fix text in `domain` column where rank is "domain" taxonomy$domain[taxonomy$rank == "domain"] <- gsub(" \\{domain\\}", "", taxonomy$fullname[taxonomy$rank == "domain"]) +# the filters above may have removed parents of records that were kept (e.g. the genus of a kept synonym), +# which would leave these records without a valid MO code later on - so add them again +n_before <- nrow(taxonomy) +taxonomy <- add_missing_parents(taxonomy, current_gbif) %>% + filter(!(rank == "kingdom" & fullname %in% c("Bacteria", "Archaea"))) +message("Added ", nrow(taxonomy) - n_before, " missing parents after filtering") +review( + taxonomy %>% filter(duplicated(fullname) | duplicated(fullname, fromLast = TRUE)), + "Duplicate full names after adding missing parents" +) +taxonomy <- taxonomy %>% + arrange(fullname, source_prio(source)) %>% + distinct(fullname, .keep_all = TRUE) %>% + # this also sets the prevalence of the added records + compute_prevalence() + # *** Save intermediate results (2b) *** ---------------------------------------------------------- @@ -2269,6 +2462,81 @@ taxonomy <- taxonomy %>% bind_rows(plasmodia) %>% arrange(fullname) +# MB 2026-10-03/ the same applies to Leishmania (in 2026, COL contained only a handful of its species, +# not L. donovani, L. major or L. aethiopica), so add the clinically relevant species that are missing +leishmania_genus <- taxonomy %>% filter(fullname == "Leishmania", rank == "genus") +if (nrow(leishmania_genus) == 1) { + leishmania <- tibble( + species = c( + "aethiopica", "amazonensis", "braziliensis", "donovani", "guyanensis", "infantum", "lainsoni", + "major", "martiniquensis", "mexicana", "naiffi", "panamensis", "peruviana", "shawi", "tropica", + "venezuelensis" + ) + ) %>% + mutate( + fullname = paste("Leishmania", species), + domain = leishmania_genus$domain, + kingdom = leishmania_genus$kingdom, + phylum = leishmania_genus$phylum, + class = leishmania_genus$class, + order = leishmania_genus$order, + family = leishmania_genus$family, + genus = "Leishmania", + subspecies = "", + rank = "species", + source = "manually added", + status = "accepted", + prevalence = leishmania_genus$prevalence + ) %>% + filter(!fullname %in% taxonomy$fullname) + message("Adding ", nrow(leishmania), " Leishmania species") + taxonomy <- taxonomy %>% + bind_rows(leishmania) %>% + arrange(fullname) +} else { + warning("Leishmania genus not found, check this!", call. = FALSE) +} + + +# Clean-up of nonsense names ---------------------------------------------------------------------- + +# (until 2026, this was applied to `microorganisms` instead of `taxonomy`, so it had no effect, and it was +# done after the MO codes were made - names such as 'Verticillium sbs.' then got a code like F_VRTCL_SBS.) +table(unlist(strsplit(taxonomy$fullname, ""))) +n_before <- nrow(taxonomy) +taxonomy <- taxonomy %>% + filter(fullname %unlike% " [0-9]") %>% + filter(fullname %unlike% " (i|ii|iii|iv|v|vi|vii|viii|ix|x)$" | genus == "Streptococcus") %>% + filter(fullname %unlike% "[.]") %>% + # ASCII names without quotes, also for the records from the previous data set + mutate(across(c(fullname, domain:subspecies), ascii_names)) %>% + distinct(fullname, .keep_all = TRUE) +message("Removed ", n_before - nrow(taxonomy), " nonsense names") + +# Fix genera that are synonyms while they contain accepted species -------------------------------- + +# e.g. in 2026, MycoBank listed the genus Blastomyces as a synonym, while Blastomyces dermatitidis was accepted +genera_with_accepted_species <- taxonomy %>% + filter(rank == "species", status == "accepted") %>% + distinct(domain, genus) +genus_synonyms_to_fix <- taxonomy %>% + filter(rank == "genus", status == "synonym") %>% + semi_join(genera_with_accepted_species, by = c("domain", "genus")) +review( + genus_synonyms_to_fix %>% select(fullname, domain, family, source, lpsn_renamed_to, mycobank_renamed_to, gbif_renamed_to), + "Synonym genera with accepted species, these will become accepted" +) +taxonomy <- taxonomy %>% + mutate( + fix = fullname %in% genus_synonyms_to_fix$fullname, + status = if_else(fix, "accepted", status), + lpsn_renamed_to = if_else(fix, NA_character_, lpsn_renamed_to), + mycobank_renamed_to = if_else(fix, NA_character_, mycobank_renamed_to), + gbif_renamed_to = if_else(fix, NA_character_, gbif_renamed_to) + ) %>% + select(-fix) +rm(genera_with_accepted_species, genus_synonyms_to_fix) + # Add microbial IDs ------------------------------------------------------------------------------- @@ -2292,8 +2560,27 @@ mo_domain <- taxonomy %>% ) mo_domain -# in 2026 we added `domain`, remove this line later and just keep it `AMR::microorganisms` -existing_mo_tbl <- AMR::microorganisms %>% rename(domain = kingdom) +# `domain` exists in the data set since 2026, so codes are kept per domain +# (in the 2026 build, this was `AMR::microorganisms %>% rename(domain = kingdom)`, which would now fail) +existing_mo_tbl <- microorganisms_old %>% + mutate(mo = as.character(mo)) + +# parts of existing MO codes, split on the underscore (genus codes can contain digits or hyphens, such as +# B_RBCLM1 or B_CND-P, so a regex with only [A-Z] does not work and caused repeated elements like +# B_SCLLM_CNNM_LNSM_LNSM_LNSM in earlier versions) +mo_part <- function(mo, i) { + vapply(strsplit(as.character(mo), "_", fixed = TRUE), function(x) if (length(x) >= i) x[i] else NA_character_, character(1)) +} +# all codes that have ever been used must never be given to another taxon, also not when their taxon +# is not in the new data set anymore (users may have stored old codes) +reserved_genus_codes <- existing_mo_tbl %>% + filter(rank %in% c("genus", "species", "subspecies")) %>% + transmute(domain, code = mo_part(mo, 2)) %>% + distinct() +reserved_species_codes <- existing_mo_tbl %>% + filter(rank %in% c("species", "subspecies")) %>% + transmute(domain, genus_code = mo_part(mo, 2), code = mo_part(mo, 3)) %>% + distinct() # phylum until family are abbreviated with 8 characters and prefixed with their rank @@ -2308,7 +2595,7 @@ mo_phylum <- taxonomy %>% transmute( domain, phylum = fullname, - mo_old = gsub("[A-Z]{1,2}_", "", as.character(mo)) + mo_old = gsub("^[A-Z]{1,2}_", "", as.character(mo)) ), by = c("domain", "phylum") ) %>% @@ -2338,7 +2625,7 @@ mo_class <- taxonomy %>% transmute( domain, class = fullname, - mo_old = gsub("[A-Z]{1,2}_", "", as.character(mo)) + mo_old = gsub("^[A-Z]{1,2}_", "", as.character(mo)) ), by = c("domain", "class") ) %>% @@ -2368,7 +2655,7 @@ mo_order <- taxonomy %>% transmute( domain, order = fullname, - mo_old = gsub("[A-Z]{1,2}_", "", as.character(mo)) + mo_old = gsub("^[A-Z]{1,2}_", "", as.character(mo)) ), by = c("domain", "order") ) %>% @@ -2398,7 +2685,7 @@ mo_family <- taxonomy %>% transmute( domain, family = fullname, - mo_old = gsub("[A-Z]{1,2}_", "", as.character(mo)) + mo_old = gsub("^[A-Z]{1,2}_", "", as.character(mo)) ), by = c("domain", "family") ) %>% @@ -2418,6 +2705,27 @@ if (any(mo_family$mo_duplicated, na.rm = TRUE)) { mo_family <- mo_family %>% select(domain, family, mo_family) +# keep the old code where available, and give every new taxon the first candidate code that is not in use +# and was never used before (`reserved`) - if all candidates are taken, a number is added to the first one +assign_codes <- function(old, candidates, reserved = character(0)) { + used <- unique(c(reserved, old[!is.na(old)])) + out <- old + for (i in which(is.na(old))) { + cand <- unique(candidates[[i]][!is.na(candidates[[i]]) & candidates[[i]] != ""]) + pick <- cand[!cand %in% used][1] + k <- 1 + while (is.na(pick)) { + if (!paste0(cand[1], k) %in% used) { + pick <- paste0(cand[1], k) + } + k <- k + 1 + } + out[i] <- pick + used <- c(used, pick) + } + out +} + # construct code part for genus - keep old code where available and generate new ones where needed mo_genus <- taxonomy %>% filter(rank == "genus") %>% @@ -2428,124 +2736,84 @@ mo_genus <- taxonomy %>% existing_mo_tbl %>% filter(rank == "genus") %>% transmute( - mo_genus_old = gsub("^[A-Z]+_", "", as.character(mo)), + mo_genus_old = mo_part(mo, 2), domain, genus ) %>% distinct(domain, genus, .keep_all = TRUE), by = c("domain", "genus") ) %>% - distinct(domain, genus, .keep_all = TRUE) %>% - # since domain is part of the code, genus abbreviations may be duplicated between kingdoms + # since domain is part of the code, genus abbreviations may be duplicated between domains group_by(domain) %>% - # generate new MO codes for genus and set the right one mutate( - mo_genus_new5 = AMR:::abbreviate_mo(genus, 5), - mo_genus_new5b = paste0(AMR:::abbreviate_mo(genus, 5), 1), - mo_genus_new5c = paste0(AMR:::abbreviate_mo(genus, 5), 2), - mo_genus_new6 = AMR:::abbreviate_mo(genus, 6), - mo_genus_new7 = AMR:::abbreviate_mo(genus, 7), - mo_genus_new8 = AMR:::abbreviate_mo(genus, 8), - mo_genus_new = case_when( - !is.na(mo_genus_old) ~ mo_genus_old, - !mo_genus_new5 %in% mo_genus_old ~ mo_genus_new5, - !mo_genus_new6 %in% mo_genus_old ~ mo_genus_new6, - !mo_genus_new7 %in% mo_genus_old ~ mo_genus_new7, - !mo_genus_new8 %in% mo_genus_old ~ mo_genus_new8, - !mo_genus_new5b %in% mo_genus_old ~ mo_genus_new5b, - TRUE ~ mo_genus_old - ), - mo_duplicated = duplicated(mo_genus_new), - mo_genus_new = case_when( - !mo_duplicated ~ mo_genus_new, - mo_duplicated & mo_genus_new == mo_genus_old ~ mo_genus_new6, - mo_duplicated & mo_genus_new == mo_genus_new5 ~ mo_genus_new6, - mo_duplicated & mo_genus_new == mo_genus_new6 ~ mo_genus_new7, - mo_duplicated & mo_genus_new == mo_genus_new7 ~ mo_genus_new8, - mo_duplicated & mo_genus_new == mo_genus_new8 ~ mo_genus_new5b, - TRUE ~ NA_character_ - ), - mo_duplicated = duplicated(mo_genus_new), - mo_genus_new = case_when( - !mo_duplicated ~ mo_genus_new, - mo_duplicated & mo_genus_new == mo_genus_new6 ~ mo_genus_new7, - mo_duplicated & mo_genus_new == mo_genus_new7 ~ mo_genus_new8, - mo_duplicated & mo_genus_new == mo_genus_new8 ~ mo_genus_new5b, - mo_duplicated ~ mo_genus_new5c, - TRUE ~ NA_character_ - ), - mo_duplicated = duplicated(mo_genus_new) + mo_genus = assign_codes( + old = mo_genus_old, + candidates = Map( + c, + AMR:::abbreviate_mo(genus, 5), + AMR:::abbreviate_mo(genus, 6), + AMR:::abbreviate_mo(genus, 7), + AMR:::abbreviate_mo(genus, 8), + paste0(AMR:::abbreviate_mo(genus, 5), 1), + paste0(AMR:::abbreviate_mo(genus, 5), 2) + ), + reserved = reserved_genus_codes$code[reserved_genus_codes$domain == cur_group()$domain] + ) ) %>% ungroup() -if (any(mo_genus$mo_duplicated, na.rm = TRUE) | anyNA(mo_genus$mo_genus_new)) { +if (anyDuplicated(paste(mo_genus$domain, mo_genus$mo_genus)) > 0 || anyNA(mo_genus$mo_genus)) { stop("Duplicate MO codes for genus!") } -# mo_genus %>% filter(mo_duplicated) -# no duplicates *within kingdoms*, so keep the right columns for left joining later +# no duplicates *within domains*, so keep the right columns for left joining later mo_genus <- mo_genus %>% - select(domain, genus, mo_genus = mo_genus_new) + select(domain, genus, mo_genus) # same for species - keep old where available and create new per domain-genus where needed: mo_species <- taxonomy %>% filter(rank == "species") %>% arrange(status, fullname) %>% # sort on accepted < synonym < unknown, added 2026-09-03 distinct(domain, genus, species) %>% + left_join(mo_genus, by = c("domain", "genus")) %>% left_join( existing_mo_tbl %>% filter(rank == "species") %>% transmute( - mo_species_old = gsub("^[A-Z]+_[A-Z]+_", "", as.character(mo)), + mo_species_old = mo_part(mo, 3), domain, genus, species ) %>% - filter(mo_species_old %unlike% "-") %>% + # only valid codes (in 2026, some old codes contained quotes, e.g. F_FUSRM_LO"ES) + filter(mo_species_old %like_case% "^[A-Z0-9]+$") %>% distinct(domain, genus, species, .keep_all = TRUE), by = c("domain", "genus", "species") ) %>% - distinct(domain, genus, species, .keep_all = TRUE) %>% group_by(domain, genus) %>% mutate( - mo_species_new4 = AMR:::abbreviate_mo(species, 4, hyphen_as_space = TRUE), - mo_species_new5 = AMR:::abbreviate_mo(species, 5, hyphen_as_space = TRUE), - mo_species_new5b = paste0( - AMR:::abbreviate_mo(species, 5, hyphen_as_space = TRUE), - 1 - ), - mo_species_new6 = AMR:::abbreviate_mo(species, 6, hyphen_as_space = TRUE), - mo_species_new7 = AMR:::abbreviate_mo(species, 7, hyphen_as_space = TRUE), - mo_species_new8 = AMR:::abbreviate_mo(species, 8, hyphen_as_space = TRUE), - mo_species_new = case_when( - !is.na(mo_species_old) ~ mo_species_old, - !mo_species_new4 %in% mo_species_old ~ mo_species_new4, - !mo_species_new5 %in% mo_species_old ~ mo_species_new5, - !mo_species_new6 %in% mo_species_old ~ mo_species_new6, - !mo_species_new7 %in% mo_species_old ~ mo_species_new7, - !mo_species_new8 %in% mo_species_old ~ mo_species_new8, - !mo_species_new5b %in% mo_species_old ~ mo_species_new5b, - TRUE ~ mo_species_old - ), - mo_duplicated = duplicated(mo_species_new), - mo_species_new = case_when( - !mo_duplicated ~ mo_species_new, - mo_duplicated & mo_species_new == mo_species_new4 ~ mo_species_new5, - mo_duplicated & mo_species_new == mo_species_new5 ~ mo_species_new6, - mo_duplicated & mo_species_new == mo_species_new6 ~ mo_species_new7, - mo_duplicated & mo_species_new == mo_species_new7 ~ mo_species_new8, - mo_duplicated & mo_species_new == mo_species_new8 ~ mo_species_new5b, - TRUE ~ NA_character_ - ), - mo_duplicated = duplicated(mo_species_new) + mo_species = assign_codes( + old = mo_species_old, + candidates = Map( + c, + AMR:::abbreviate_mo(species, 4, hyphen_as_space = TRUE), + AMR:::abbreviate_mo(species, 5, hyphen_as_space = TRUE), + AMR:::abbreviate_mo(species, 6, hyphen_as_space = TRUE), + AMR:::abbreviate_mo(species, 7, hyphen_as_space = TRUE), + AMR:::abbreviate_mo(species, 8, hyphen_as_space = TRUE), + paste0(AMR:::abbreviate_mo(species, 5, hyphen_as_space = TRUE), 1) + ), + reserved = reserved_species_codes$code[ + reserved_species_codes$domain == cur_group()$domain & + reserved_species_codes$genus_code %in% mo_genus[1] + ] + ) ) %>% ungroup() -if ( - any(mo_species$mo_duplicated, na.rm = TRUE) | anyNA(mo_species$mo_species_new) -) { +if (anyDuplicated(paste(mo_species$domain, mo_species$genus, mo_species$mo_species)) > 0 || anyNA(mo_species$mo_species)) { stop("Duplicate MO codes for species!") } -# no duplicates *within kingdoms*, so keep the right columns for left joining later +# no duplicates *within genera*, so keep the right columns for left joining later mo_species <- mo_species %>% - select(domain, genus, species, mo_species = mo_species_new) + select(domain, genus, species, mo_species) # same for subspecies - keep old where available and create new per domain-genus-species where needed: mo_subspecies <- taxonomy %>% @@ -2556,89 +2824,40 @@ mo_subspecies <- taxonomy %>% existing_mo_tbl %>% filter(rank %in% c("subspecies", "subsp.", "infraspecies")) %>% transmute( - mo_subspecies_old = gsub( - "^[A-Z]+_[A-Z]+_[A-Z]+_", - "", - as.character(mo) - ), + mo_subspecies_old = mo_part(mo, 4), domain, genus, species, subspecies ) %>% - filter(mo_subspecies_old %unlike% "-") %>% + # only valid codes (in 2026, some old codes contained quotes, e.g. F_FUSRM_LO"ES) + filter(mo_subspecies_old %like_case% "^[A-Z0-9]+$") %>% distinct(domain, genus, species, subspecies, .keep_all = TRUE), by = c("domain", "genus", "species", "subspecies") ) %>% - distinct(domain, genus, species, subspecies, .keep_all = TRUE) %>% group_by(domain, genus, species) %>% mutate( - mo_subspecies_new4 = AMR:::abbreviate_mo( - subspecies, - 4, - hyphen_as_space = TRUE - ), - mo_subspecies_new5 = AMR:::abbreviate_mo( - subspecies, - 5, - hyphen_as_space = TRUE - ), - mo_subspecies_new5b = paste0( - AMR:::abbreviate_mo(subspecies, 5, hyphen_as_space = TRUE), - 1 - ), - mo_subspecies_new6 = AMR:::abbreviate_mo( - subspecies, - 6, - hyphen_as_space = TRUE - ), - mo_subspecies_new7 = AMR:::abbreviate_mo( - subspecies, - 7, - hyphen_as_space = TRUE - ), - mo_subspecies_new8 = AMR:::abbreviate_mo( - subspecies, - 8, - hyphen_as_space = TRUE - ), - mo_subspecies_new = case_when( - !is.na(mo_subspecies_old) ~ mo_subspecies_old, - !mo_subspecies_new4 %in% mo_subspecies_old ~ mo_subspecies_new4, - !mo_subspecies_new5 %in% mo_subspecies_old ~ mo_subspecies_new5, - !mo_subspecies_new6 %in% mo_subspecies_old ~ mo_subspecies_new6, - !mo_subspecies_new7 %in% mo_subspecies_old ~ mo_subspecies_new7, - !mo_subspecies_new8 %in% mo_subspecies_old ~ mo_subspecies_new8, - !mo_subspecies_new5b %in% mo_subspecies_old ~ mo_subspecies_new5b, - TRUE ~ mo_subspecies_old - ), - mo_duplicated = duplicated(mo_subspecies_new), - mo_subspecies_new = case_when( - !mo_duplicated ~ mo_subspecies_new, - mo_duplicated & mo_subspecies_new == mo_subspecies_new4 ~ - mo_subspecies_new5, - mo_duplicated & mo_subspecies_new == mo_subspecies_new5 ~ - mo_subspecies_new6, - mo_duplicated & mo_subspecies_new == mo_subspecies_new6 ~ - mo_subspecies_new7, - mo_duplicated & mo_subspecies_new == mo_subspecies_new7 ~ - mo_subspecies_new8, - mo_duplicated & mo_subspecies_new == mo_subspecies_new8 ~ - mo_subspecies_new5b, - TRUE ~ NA_character_ - ), - mo_duplicated = duplicated(mo_subspecies_new) + mo_subspecies = assign_codes( + old = mo_subspecies_old, + candidates = Map( + c, + AMR:::abbreviate_mo(subspecies, 4, hyphen_as_space = TRUE), + AMR:::abbreviate_mo(subspecies, 5, hyphen_as_space = TRUE), + AMR:::abbreviate_mo(subspecies, 6, hyphen_as_space = TRUE), + AMR:::abbreviate_mo(subspecies, 7, hyphen_as_space = TRUE), + AMR:::abbreviate_mo(subspecies, 8, hyphen_as_space = TRUE), + paste0(AMR:::abbreviate_mo(subspecies, 5, hyphen_as_space = TRUE), 1) + ) + ) ) %>% ungroup() -if ( - any(mo_subspecies$mo_duplicated, na.rm = TRUE) | - anyNA(mo_subspecies$mo_subspecies_new) -) { +if (anyDuplicated(paste(mo_subspecies$domain, mo_subspecies$genus, mo_subspecies$species, mo_subspecies$mo_subspecies)) > 0 || + anyNA(mo_subspecies$mo_subspecies)) { stop("Duplicate MO codes for subspecies!") } -# no duplicates *within kingdoms*, so keep the right columns for left joining later +# no duplicates *within species*, so keep the right columns for left joining later mo_subspecies <- mo_subspecies %>% - select(domain, genus, species, subspecies, mo_subspecies = mo_subspecies_new) + select(domain, genus, species, subspecies, mo_subspecies) # unknowns - manually added mo_unknown <- existing_mo_tbl %>% @@ -2686,7 +2905,7 @@ taxonomy <- taxonomy %>% # such as Kapibacteria (both genus and class), but the class last time correctly got the " {class}" suffix in the fullname taxonomy %>% filter(fullname %in% .[duplicated(fullname), "fullname", drop = TRUE]) %>% - View() + review("Duplicate full names after adding MO codes") # OTHERWISE uncomment this section: # # we will prefer domains as Bacteria > Fungi > Protozoa > Archaea > Animalia, so check the ones within 1 domain @@ -2727,9 +2946,10 @@ taxonomy$mo[taxonomy$source == "manually added"][!is.na(manual_mos)] <- manual_m # filter(!(mo %like% "__" & source == "manually added")) # this must not exist (2024: GBIF mess with missing genera - remove them): +# (since 2026, missing parents are added again after filtering, so this should be 0 rows) taxonomy %>% filter(mo %like% "__") %>% - View() + review("Records without a valid MO code, these will be removed") taxonomy <- taxonomy %>% filter(mo %unlike% "__") # keep unique taxonomy @@ -2750,7 +2970,7 @@ saveRDS(taxonomy, "data-raw/taxonomy2c.rds") taxonomy %>% filter(mo %in% .[duplicated(mo), "mo", drop = TRUE]) %>% mutate(removed_if_distinct_is_applied = duplicated(mo), .before = 1) %>% - View() + review("Duplicate MO codes") # checked 2026, can all be removed, they are actual duplicates taxonomy <- taxonomy %>% distinct(mo, .keep_all = TRUE) @@ -2770,7 +2990,7 @@ taxonomy <- taxonomy %>% taxonomy %>% arrange(fullname) %>% filter(fullname %in% .[duplicated(fullname), "fullname", drop = TRUE]) %>% - View() + review("Duplicate full names") # # we can now solve it like this, but check every year if this is okay # taxonomy <- taxonomy %>% # group_by(fullname) %>% @@ -2919,17 +3139,35 @@ message( # these are the new ones: taxonomy %>% - # filter(!paste(kingdom, fullname) %in% paste(AMR::microorganisms$kingdom, AMR::microorganisms$fullname)) %>% - filter(!fullname %in% AMR::microorganisms$fullname) %>% - View() + filter(!fullname %in% microorganisms_old$fullname) %>% + review("New taxa") # these are to be removed: -AMR::microorganisms %>% +microorganisms_old %>% filter(!fullname %in% taxonomy$fullname) %>% - View() + review("Removed taxa") # be extra critical to these: -AMR::microorganisms %>% +microorganisms_old %>% filter(!fullname %in% taxonomy$fullname, source == "LPSN", status == "accepted") %>% count(rank) +# and to these: clinically relevant taxa that will be removed +microorganisms_old %>% + filter(!fullname %in% taxonomy$fullname, prevalence < 2, rank %in% c("genus", "species")) %>% + select(mo, fullname, status, source, prevalence) %>% + review("Removed taxa that were clinically relevant (prevalence < 2)") +# and these: taxa that moved to another domain (this changes their MO code) +taxonomy %>% + select(fullname, domain, mo) %>% + inner_join(microorganisms_old %>% select(fullname, domain_old = domain, mo_old = mo), by = "fullname") %>% + filter(domain != domain_old) %>% + count(domain_old, domain, genus = gsub(" .*", "", fullname), sort = TRUE) %>% + review("Genera that moved to another domain") +# and the family of the most relevant genera, which should be checked every year +taxonomy %>% + filter(rank == "genus", prevalence <= 1.25) %>% + select(fullname, domain, phylum, class, order, family, source, status) %>% + left_join(microorganisms_old %>% select(fullname, family_old = family), by = "fullname") %>% + filter(family != family_old | family == "") %>% + review("Relevant genera with a new or empty family") # Add SNOMED CT ----------------------------------------------------------------------------------- @@ -2981,12 +3219,15 @@ bacdive <- vroom::vroom(file_oxygen_tolerance, skip = 2) %>% select(species, oxygen = `Oxygen tolerance`) bacdive <- bacdive %>% # fill in missing species from previous rows - mutate(fullname = if_else(is.na(species), lag(species), species)) %>% + # a species can have multiple rows, the next rows have an empty species (lag() only filled one row and + # was then removed by the filter below anyway, so all those values were lost) + mutate(fullname = species) %>% + tidyr::fill(fullname, .direction = "down") %>% filter( - !is.na(species), + !is.na(fullname), !is.na(oxygen), oxygen %unlike% "tolerant", - species %unlike% "unclassified" + fullname %unlike% "unclassified" ) %>% select(-species) bacdive <- bacdive %>% @@ -3086,11 +3327,14 @@ bacdive_shape <- vroom::vroom(file_cell_shape, skip = 2, guess_max = 1e5) %>% select(species, shape = `Cell shape`) bacdive_shape <- bacdive_shape %>% # fill in missing species from previous rows - mutate(fullname = if_else(is.na(species), lag(species), species)) %>% + # a species can have multiple rows, the next rows have an empty species (lag() only filled one row and + # was then removed by the filter below anyway, so all those values were lost) + mutate(fullname = species) %>% + tidyr::fill(fullname, .direction = "down") %>% filter( - !is.na(species), + !is.na(fullname), !is.na(shape), - species %unlike% "unclassified" + fullname %unlike% "unclassified" ) %>% select(-species) bacdive_shape <- bacdive_shape %>% @@ -3291,17 +3535,21 @@ taxonomy$rank[which(taxonomy$fullname %like% "unknown")] <- "(unknown rank)" # this happened in early 2025 (and May 2026), check that MO codes do not have repeated elements # fixed it then like this: microorganisms$mo <- gsub("B_SCLLM_CNNM_LNSM_LNSM_LNSM_LNSM", "B_SCLLM_CNNM", microorganisms$mo) +# (the cause was found in October 2026: the old species code was extracted with a regex that failed on genus +# codes with digits, such as B_RBCLM1 - this is now done with mo_part(), so this should be 0 rows) taxonomy %>% filter(mo %like% "_.*_.*_.*_") %>% - View() + review("MO codes with repeated elements") # Pattern: capture the prefix, then remove its repetition before the final epithet -taxonomy$mo[taxonomy$mo %like% "_.*_.*_.*_"] <- gsub("^(B_[A-Z0-9]+(?:_[A-Z0-9]+)?)_\\1_([A-Z]+)$", - "\\1_\\2", - taxonomy$mo[taxonomy$mo %like% "_.*_.*_.*_"], - perl = TRUE) +taxonomy$mo[taxonomy$mo %like% "_.*_.*_.*_"] <- gsub( + "^([A-Z]{1,2}_[A-Z0-9-]+(?:_[A-Z0-9-]+)?)_\\1_([A-Z0-9-]+)$", + "\\1_\\2", + taxonomy$mo[taxonomy$mo %like% "_.*_.*_.*_"], + perl = TRUE +) taxonomy %>% filter(mo %like% "_.*_.*_.*_") %>% - View() + review("MO codes with repeated elements after fixing") # Remove childless non-bacterial genera ----------------------------------------------------------- @@ -3327,38 +3575,36 @@ for (rank_name in c("family", "order", "class", "phylum", "kingdom")) { message("Removed ", n_before - nrow(taxonomy), " childless ", rank_name, " entries") } -# Clean-up of nonsense names ---------------------------------------------------------------------- - -table(unlist(strsplit(microorganisms$fullname, ""))) -microorganisms <- microorganisms %>% - filter(fullname %unlike% " [0-9]") %>% - filter(fullname %unlike% " (i|ii|iii|iv|v|vi|vii|viii|ix|x)$" | genus == "Streptococcus") %>% - filter(fullname %unlike% "[.]") %>% - mutate(across(c(fullname, domain:subspecies), function(x) gsub("[\"'`]", "", x))) - +# records that were only inferred by this script get the same label as before +taxonomy$source[taxonomy$source == "inferred"] <- "manually added" # Some final checks ------------------------------------------------------------------------------- -# all our previously manually added names should be in it -all( - microorganisms$fullname[microorganisms$source == "manually added"] %in% - taxonomy$fullname -) -microorganisms$fullname[ - !microorganisms$fullname[microorganisms$source == "manually added"] %in% - taxonomy$fullname -] - -# we added an MO code, so make sure everything is still unique -any(duplicated(taxonomy$mo)) -any(duplicated(taxonomy$fullname)) +# all our previously manually added names should be in it - review the ones that are not +# (until 2026, also records that were only inferred were labelled 'manually added', so many will be gone) +microorganisms_old %>% + filter(source == "manually added", !fullname %in% taxonomy$fullname) %>% + select(mo, fullname, rank, status, prevalence) %>% + review("Previously manually added taxa that are not in the new data set") # LPSN records must be prokaryotes and MycoBank records must be fungi, also in their MO code (issue #309) -all(taxonomy$domain[taxonomy$source == "LPSN"] %in% c("Bacteria", "Archaea")) -all(taxonomy$mo[taxonomy$source == "LPSN"] %like% "^(B|A)_") -all(taxonomy$domain[taxonomy$source == "MycoBank"] == "Fungi") -all(taxonomy$mo[taxonomy$source == "MycoBank"] %like% "^F_") +if (!all(taxonomy$domain[taxonomy$source == "LPSN"] %in% c("Bacteria", "Archaea")) || + !all(taxonomy$mo[taxonomy$source == "LPSN"] %like% "^(B|A)_") || + !all(taxonomy$domain[taxonomy$source == "MycoBank"] == "Fungi") || + !all(taxonomy$mo[taxonomy$source == "MycoBank"] %like% "^F_")) { + warning("LPSN records outside prokaryotes or MycoBank records outside fungi, check 'Combine the datasets'", call. = FALSE) +} + +# these must never fail, as the package relies on them +stopifnot( + "MO codes must be unique" = !anyDuplicated(taxonomy$mo), + "full names must be unique" = !anyDuplicated(taxonomy$fullname), + "MO codes must not be empty" = !any(taxonomy$mo %in% c("", NA)), + "prevalence must not be empty" = !anyNA(taxonomy$prevalence), + "status must be accepted, synonym or unknown" = all(taxonomy$status %in% c("accepted", "synonym", "unknown")), + "sources must be known" = all(taxonomy$source %in% c("LPSN", "MycoBank", "GBIF", "manually added")) +) # Update other data sets -------------------------------------------------------------------------- @@ -3366,7 +3612,7 @@ all(taxonomy$mo[taxonomy$source == "MycoBank"] %like% "^F_") fix_old_mos <- function(dataset, new_ref, drop = FALSE, col = "mo") { before <- dataset - mo_names <- AMR::microorganisms$fullname[match(before[[col]], AMR::microorganisms$mo)] + mo_names <- microorganisms_old$fullname[match(before[[col]], microorganisms_old$mo)] matches <- new_ref$mo[match(mo_names, new_ref$fullname)] after <- before after[[col]] <- matches @@ -3429,6 +3675,13 @@ taxonomy <- taxonomy %>% ) %>% AMR:::dataset_UTF8_to_ASCII() +# check again after the conversion to ASCII +stopifnot( + "full names must be unique after conversion to ASCII" = !anyDuplicated(taxonomy$fullname), + "MO codes must be unique after conversion to ASCII" = !anyDuplicated(taxonomy$mo), + "MO codes may only contain A-Z, 0-9, _, -, [ and ]" = all(taxonomy$mo %like_case% "^[][A-Z0-9_-]+$") +) + microorganisms <- taxonomy # set class diff --git a/data-raw/_reproduction_scripts/reproduction_of_microorganisms_CLAUDE.md b/data-raw/_reproduction_scripts/reproduction_of_microorganisms_CLAUDE.md new file mode 100644 index 000000000..1bdbb2897 --- /dev/null +++ b/data-raw/_reproduction_scripts/reproduction_of_microorganisms_CLAUDE.md @@ -0,0 +1,284 @@ +# Unattended update of the `microorganisms` data set + +You are updating the taxonomy of the AMR package for R: the `microorganisms` data set, built by +`data-raw/_reproduction_scripts/reproduction_of_microorganisms.R`. Hundreds of thousands of people +rely on this data set, in clinical microbiology, surveillance and research. An error here, such as a +pathogen in the wrong kingdom, a missing genus or an MO code that changes meaning, propagates into +patient-related analyses worldwide. Work like a senior clinical microbiologist who is also a +meticulous data engineer: careful, sceptical of every source, and never satisfied with "it ran +without errors". + +Work autonomously from start to finish. Only stop early for the reasons listed under +"When to stop". Everything you do ends up in one new pull request. + +**A human is in the loop.** You do the work, a human decides. Every finding, every decision you +took and every point you could not resolve must be written to the repository (git-tracked) and to the +pull request as a checklist, so that the human can review each point separately before merging. You +never merge the PR yourself. In the text below, `YYYY-MM` is the year and month of this update, and +`` is `data-raw/taxonomy_review/YYYY-MM`. + +## Non-negotiable rules + +- **Never download the source files yourself.** GBIF/COL, LPSN, MycoBank and BacDive require + registration or manual export buttons, so a human downloads them. If the script stops because + files are missing, report exactly which files are missing, with the steps at the top of the + script, and stop. +- **Never edit `data/*.rda` by hand** and never patch individual records afterwards. Every fix goes + into the script, so that next year's run includes it. The script is the product, the data set is + its output. +- **Fix root causes, not symptoms.** If a genus has the wrong family, find out why (source data, + parser, priority rule, homonym) and fix that mechanism. Only use the explicit override lists in + the script (`genus_domain_override`, the Plasmodium and Leishmania blocks) when the cause is a + genuine flaw in the source data that no rule can resolve, and explain each addition in a code comment. +- **Never fabricate.** No invented species names, citations, release dates or test results. If you + cannot verify something, say so in the PR. +- **Never commit to or push to `main`.** Work on a new branch and open a new PR. +- Follow `CLAUDE.md` in the repository root: tidyverse style, `%like%`/`%like_case%` instead of + `grepl()`, British English, zero dependencies (packages used only in `data-raw/` are fine). +- Never write to or modify anything under `/etc`. + +## Step 0: preflight + +1. Check the working tree is clean, `git fetch --tags` and update `main`, then create a branch + `taxonomy-update-YYYY-MM` (current year and month) from `origin/main`. +2. Check resources: at least 16 GB RAM (`free -g`) and 10 GB free disk space. If not, stop. +3. Determine the folder with the raw files. Use the environment variable `AMR_TAXONOMY_SOURCE_DIR` if + set, otherwise `~/Downloads/`. Run only the guard at the top of the script, from + `folder_location <- ...` up to and including `rm(raw_missing, raw_downloads, raw_age)`, in a + separate R process. If it stops, report and stop. If it warns about old files, compare the + dates with the release dates (step 1) before continuing. If the files are clearly from the + previous update, stop and report. +4. Keep the previous release for comparison: + `git show origin/main:data/microorganisms.rda > /microorganisms_old.rda`, where + `` is a folder outside the repository. + +## Step 1: record the source versions + +These are needed for `TAXONOMY_VERSION` in `R/aa_globals.R` later. Write them down now: + +- **GBIF/COL**: read `eml.xml` (or `metadata.yaml`) in the folder of `Taxon.tsv` for the release + title, date and DOI. The citation format must match the existing one in `R/aa_globals.R`. +- **LPSN, MycoBank, BacDive**: the accessed date is the modification date of the downloaded file. +- Do not guess any of these. If a value cannot be found, leave the existing value and list it in the + PR under "Needs a human". + +## Step 2: run the build + +The script takes several hours (most of it is LPSN scraping, which is cached in +`data-raw/lpsn_scrape_cache.rds`, so a restart is fast). Run it detached from your tool calls, so that +it survives time limits, and write everything to a log: + +```bash +mkdir -p /logs +cd +nohup Rscript -e 'options(AMR_build_view = FALSE, AMR_build_print_rows = 200, AMR_build_review_dir = "", warn = 1, width = 200); source("data-raw/_reproduction_scripts/reproduction_of_microorganisms.R", echo = TRUE, max.deparse.length = Inf, keep.source = TRUE)' > /logs/build_1.log 2>&1 & +``` + +Monitor the log at sensible intervals (every 10 to 20 minutes, not continuously). Read every +`>>> REVIEW:` block as soon as it appears; you do not have to wait until the end to start thinking. +With `AMR_build_review_dir` set, every review block is also saved in full as a CSV file in +``, named after its title. These files are part of the PR. Set the same option in every +resumed run. + +### Intermediate results + +The script saves its state at fixed points, exactly as in previous years. Keep this structure intact +and do not add, rename or remove checkpoints without a clear reason: + +| File in `data-raw/` | Saved after | +|---|---| +| `taxonomy_lpsn.rds`, `taxonomy_mycobank.rds`, `taxonomy_gbif.rds` | reading each source | +| `taxonomy_lpsn_missing.rds` | scraping the LPSN lineage of all genera | +| `taxonomy0.rds` | combining the sources and resolving homonyms | +| `taxonomy1.rds` | adding missing parents and deduplication | +| `taxonomy1b.rds` | LPSN lookup of GBIF records | +| `taxonomy1c.rds` | fixing GBIF taxonomy with LPSN/MycoBank | +| `taxonomy2.rds` | prevalence | +| `taxonomy2b.rds` | removing unwanted records and harmonising taxonomy | +| `taxonomy2c.rds` | MO codes | +| `taxonomy3.rds` | integrity checks and parent identifiers | +| `taxonomy3b.rds` | all additions, just before saving to the package | + +These files are tracked in git and must be committed in the PR, so that the human can inspect every +stage. The same applies to everything in ``. `lpsn_scrape_cache.rds` is ignored by git +and must not be committed. + +### If the run fails + +1. Read the error and the 100 lines before it. Find the root cause in the script, fix it, and + check the syntax with `Rscript -e 'invisible(parse("