mirror of https://github.com/msberends/AMR.git
499 lines
21 KiB
R
499 lines
21 KiB
R
# ==================================================================== #
|
|
# TITLE #
|
|
# Antimicrobial Resistance (AMR) Data Analysis for R #
|
|
# #
|
|
# SOURCE #
|
|
# https://github.com/msberends/AMR #
|
|
# #
|
|
# LICENCE #
|
|
# (c) 2018-2022 Berends MS, Luz CF et al. #
|
|
# Developed at the University of Groningen, the Netherlands, in #
|
|
# collaboration with non-profit organisations Certe Medical #
|
|
# Diagnostics & Advice, and University Medical Center Groningen. #
|
|
# #
|
|
# This R package is free software; you can freely use and distribute #
|
|
# it for both personal and commercial purposes under the terms of the #
|
|
# GNU General Public License version 2.0 (GNU GPL-2), as published by #
|
|
# the Free Software Foundation. #
|
|
# We created this package for both routine data analysis and academic #
|
|
# research and it was publicly released in the hope that it will be #
|
|
# useful, but it comes WITHOUT ANY WARRANTY OR LIABILITY. #
|
|
# #
|
|
# Visit our website for the full manual and a complete tutorial about #
|
|
# how to conduct AMR data analysis: https://msberends.github.io/AMR/ #
|
|
# ==================================================================== #
|
|
|
|
# Run this file to update the package using:
|
|
# source("data-raw/_pre_commit_hook.R")
|
|
|
|
library(dplyr, warn.conflicts = FALSE)
|
|
devtools::load_all(quiet = TRUE)
|
|
|
|
suppressMessages(set_AMR_locale("English"))
|
|
|
|
old_globalenv <- ls(envir = globalenv())
|
|
|
|
# Save internal data to R/sysdata.rda -------------------------------------
|
|
|
|
# See 'data-raw/eucast_rules.tsv' for the EUCAST reference file
|
|
EUCAST_RULES_DF <- utils::read.delim(
|
|
file = "data-raw/eucast_rules.tsv",
|
|
skip = 10,
|
|
sep = "\t",
|
|
stringsAsFactors = FALSE,
|
|
header = TRUE,
|
|
strip.white = TRUE,
|
|
na = c(NA, "", NULL)
|
|
) %>%
|
|
# take the order of the reference.rule_group column in the original data file
|
|
mutate(
|
|
reference.rule_group = factor(reference.rule_group,
|
|
levels = unique(reference.rule_group),
|
|
ordered = TRUE
|
|
),
|
|
sorting_rule = ifelse(grepl("^Table", reference.rule, ignore.case = TRUE), 1, 2)
|
|
) %>%
|
|
arrange(
|
|
reference.rule_group,
|
|
reference.version,
|
|
sorting_rule,
|
|
reference.rule
|
|
) %>%
|
|
mutate(reference.rule_group = as.character(reference.rule_group)) %>%
|
|
select(-sorting_rule)
|
|
|
|
TRANSLATIONS <- utils::read.delim(
|
|
file = "data-raw/translations.tsv",
|
|
sep = "\t",
|
|
stringsAsFactors = FALSE,
|
|
header = TRUE,
|
|
blank.lines.skip = TRUE,
|
|
fill = TRUE,
|
|
strip.white = TRUE,
|
|
encoding = "UTF-8",
|
|
fileEncoding = "UTF-8",
|
|
na.strings = c(NA, "", NULL),
|
|
allowEscapes = TRUE, # else "\\1" will be imported as "\\\\1"
|
|
quote = ""
|
|
)
|
|
|
|
LANGUAGES_SUPPORTED_NAMES <- c(
|
|
list(en = list(exonym = "English", endonym = "English")),
|
|
lapply(
|
|
TRANSLATIONS[, which(nchar(colnames(TRANSLATIONS)) == 2), drop = FALSE],
|
|
function(x) list(exonym = x[1], endonym = x[2])
|
|
)
|
|
)
|
|
|
|
LANGUAGES_SUPPORTED <- names(LANGUAGES_SUPPORTED_NAMES)
|
|
|
|
# vectors of CoNS and CoPS, improves speed in as.mo()
|
|
create_species_cons_cops <- function(type = c("CoNS", "CoPS")) {
|
|
# Determination of which staphylococcal species are CoNS/CoPS according to:
|
|
# - Becker et al. 2014, PMID 25278577
|
|
# - Becker et al. 2019, PMID 30872103
|
|
# - Becker et al. 2020, PMID 32056452
|
|
# this function returns class <mo>
|
|
MO_staph <- AMR::microorganisms
|
|
MO_staph <- MO_staph[which(MO_staph$genus == "Staphylococcus"), , drop = FALSE]
|
|
if (type == "CoNS") {
|
|
MO_staph[which(MO_staph$species %in% c(
|
|
"coagulase-negative", "argensis", "arlettae",
|
|
"auricularis", "borealis", "caeli", "capitis", "caprae",
|
|
"carnosus", "casei", "chromogenes", "cohnii", "condimenti",
|
|
"croceilyticus",
|
|
"debuckii", "devriesei", "edaphicus", "epidermidis",
|
|
"equorum", "felis", "fleurettii", "gallinarum",
|
|
"haemolyticus", "hominis", "jettensis", "kloosii",
|
|
"lentus", "lugdunensis", "massiliensis", "microti",
|
|
"muscae", "nepalensis", "pasteuri", "petrasii",
|
|
"pettenkoferi", "piscifermentans", "pragensis", "pseudoxylosus",
|
|
"pulvereri", "rostri", "saccharolyticus", "saprophyticus",
|
|
"sciuri", "simulans", "stepanovicii", "succinus",
|
|
"ureilyticus",
|
|
"vitulinus", "vitulus", "warneri", "xylosus",
|
|
"caledonicus", "canis",
|
|
"durrellii", "lloydii",
|
|
"ratti", "taiwanensis"
|
|
) |
|
|
# old, now renamed to S. schleiferi (but still as synonym in our data of course):
|
|
(MO_staph$species == "schleiferi" & MO_staph$subspecies %in% c("schleiferi", ""))),
|
|
"mo",
|
|
drop = TRUE
|
|
]
|
|
} else if (type == "CoPS") {
|
|
MO_staph[which(MO_staph$species %in% c(
|
|
"coagulase-positive", "coagulans",
|
|
"agnetis", "argenteus",
|
|
"cornubiensis",
|
|
"delphini", "lutrae",
|
|
"hyicus", "intermedius",
|
|
"pseudintermedius", "pseudointermedius",
|
|
"schweitzeri", "simiae",
|
|
"roterodami",
|
|
"singaporensis"
|
|
) |
|
|
# old, now renamed to S. coagulans (but still as synonym in our data of course):
|
|
(MO_staph$species == "schleiferi" & MO_staph$subspecies == "coagulans")),
|
|
"mo",
|
|
drop = TRUE
|
|
]
|
|
}
|
|
}
|
|
create_MO_fullname_lower <- function() {
|
|
MO_lookup <- AMR::microorganisms
|
|
# use this paste instead of `fullname` to work with Viridans Group Streptococci, etc.
|
|
MO_lookup$fullname_lower <- tolower(trimws(paste(
|
|
MO_lookup$genus,
|
|
MO_lookup$species,
|
|
MO_lookup$subspecies
|
|
)))
|
|
ind <- MO_lookup$genus == "" | grepl("^[(]unknown ", MO_lookup$fullname, perl = TRUE)
|
|
MO_lookup[ind, "fullname_lower"] <- tolower(MO_lookup[ind, "fullname", drop = TRUE])
|
|
MO_lookup$fullname_lower <- trimws(gsub("[^.a-z0-9/ \\-]+", "", MO_lookup$fullname_lower, perl = TRUE))
|
|
MO_lookup$fullname_lower
|
|
}
|
|
MO_CONS <- create_species_cons_cops("CoNS")
|
|
MO_COPS <- create_species_cons_cops("CoPS")
|
|
MO_STREP_ABCG <- MO_lookup$mo[which(MO_lookup$genus == "Streptococcus" &
|
|
MO_lookup$species %in% c(
|
|
"pyogenes", "agalactiae", "dysgalactiae", "equi", "anginosus", "sanguinis", "salivarius",
|
|
"group A", "group B", "group C", "group D", "group F", "group G", "group H", "group K", "group L"
|
|
))]
|
|
MO_FULLNAME_LOWER <- create_MO_fullname_lower()
|
|
MO_PREVALENT_GENERA <- c(
|
|
"Absidia", "Acanthamoeba", "Acholeplasma", "Acremonium", "Actinotignum", "Aedes", "Alistipes", "Alloprevotella",
|
|
"Alternaria", "Amoeba", "Anaerosalibacter", "Ancylostoma", "Angiostrongylus", "Anisakis", "Anopheles",
|
|
"Apophysomyces", "Arachnia", "Aspergillus", "Aureobasidium", "Bacteroides", "Basidiobolus",
|
|
"Beauveria", "Bergeyella", "Blastocystis", "Blastomyces", "Borrelia", "Brachyspira", "Branhamella",
|
|
"Butyricimonas", "Candida", "Capillaria", "Capnocytophaga", "Catabacter", "Cetobacterium", "Chaetomium",
|
|
"Chlamydia", "Chlamydophila", "Chryseobacterium", "Chrysonilia", "Cladophialophora", "Cladosporium",
|
|
"Conidiobolus", "Contracaecum", "Cordylobia", "Cryptococcus", "Curvularia", "Deinococcus", "Demodex",
|
|
"Dermatobia", "Dientamoeba", "Diphyllobothrium", "Dirofilaria", "Dysgonomonas", "Echinostoma", "Elizabethkingia",
|
|
"Empedobacter", "Entamoeba", "Enterobius", "Exophiala", "Exserohilum", "Fasciola", "Flavobacterium", "Fonsecaea",
|
|
"Fusarium", "Fusobacterium", "Giardia", "Haloarcula", "Halobacterium", "Halococcus", "Hendersonula",
|
|
"Heterophyes", "Histomonas", "Histoplasma", "Hymenolepis", "Hypomyces", "Hysterothylacium", "Leishmania", "Lelliottia",
|
|
"Leptosphaeria", "Leptotrichia", "Lucilia", "Lumbricus", "Malassezia", "Malbranchea", "Metagonimus",
|
|
"Microsporidium", "Microsporum", "Mortierella", "Mucor", "Mycocentrospora", "Mycoplasma", "Myroides", "Necator",
|
|
"Nectria", "Ochroconis", "Odoribacter", "Oesophagostomum", "Oidiodendron", "Opisthorchis",
|
|
"Ornithobacterium", "Parabacteroides", "Pediculus", "Pedobacter", "Phlebotomus", "Phocaeicola",
|
|
"Phocanema", "Phoma", "Piedraia", "Pithomyces", "Pityrosporum", "Porphyromonas", "Prevotella",
|
|
"Pseudallescheria", "Pseudoterranova", "Pulex", "Rhizomucor", "Rhizopus", "Rhodotorula", "Riemerella",
|
|
"Saccharomyces", "Sarcoptes", "Scolecobasidium", "Scopulariopsis", "Scytalidium", "Sphingobacterium",
|
|
"Spirometra", "Spiroplasma", "Sporobolomyces", "Stachybotrys", "Streptobacillus", "Strongyloides",
|
|
"Syngamus", "Taenia", "Tannerella", "Tenacibaculum", "Terrimonas", "Toxocara", "Treponema", "Trichinella",
|
|
"Trichobilharzia", "Trichoderma", "Trichomonas", "Trichophyton", "Trichosporon", "Trichostrongylus",
|
|
"Trichuris", "Tritirachium", "Trypanosoma", "Trombicula", "Tunga", "Ureaplasma", "Victivallis", "Wautersiella",
|
|
"Weeksella", "Wuchereria"
|
|
)
|
|
|
|
# antibiotic groups
|
|
# (these will also be used for eucast_rules() and understanding data-raw/eucast_rules.tsv)
|
|
globalenv_before_ab <- c(ls(envir = globalenv()), "globalenv_before_ab")
|
|
AB_AMINOGLYCOSIDES <- antibiotics %>%
|
|
filter(group %like% "aminoglycoside") %>%
|
|
pull(ab)
|
|
AB_AMINOPENICILLINS <- as.ab(c("AMP", "AMX"))
|
|
AB_ANTIFUNGALS <- AB_lookup %>%
|
|
filter(group %like% "antifungal") %>%
|
|
pull(ab)
|
|
AB_ANTIMYCOBACTERIALS <- AB_lookup %>%
|
|
filter(group %like% "antimycobacterial") %>%
|
|
pull(ab)
|
|
AB_CARBAPENEMS <- antibiotics %>%
|
|
filter(group %like% "carbapenem") %>%
|
|
pull(ab)
|
|
AB_CEPHALOSPORINS <- antibiotics %>%
|
|
filter(group %like% "cephalosporin") %>%
|
|
pull(ab)
|
|
AB_CEPHALOSPORINS_1ST <- antibiotics %>%
|
|
filter(group %like% "cephalosporin.*1") %>%
|
|
pull(ab)
|
|
AB_CEPHALOSPORINS_2ND <- antibiotics %>%
|
|
filter(group %like% "cephalosporin.*2") %>%
|
|
pull(ab)
|
|
AB_CEPHALOSPORINS_3RD <- antibiotics %>%
|
|
filter(group %like% "cephalosporin.*3") %>%
|
|
pull(ab)
|
|
AB_CEPHALOSPORINS_4TH <- antibiotics %>%
|
|
filter(group %like% "cephalosporin.*4") %>%
|
|
pull(ab)
|
|
AB_CEPHALOSPORINS_5TH <- antibiotics %>%
|
|
filter(group %like% "cephalosporin.*5") %>%
|
|
pull(ab)
|
|
AB_CEPHALOSPORINS_EXCEPT_CAZ <- AB_CEPHALOSPORINS[AB_CEPHALOSPORINS != "CAZ"]
|
|
AB_FLUOROQUINOLONES <- antibiotics %>%
|
|
filter(atc_group2 %like% "fluoroquinolone" | (group %like% "quinolone" & is.na(atc_group2))) %>%
|
|
pull(ab)
|
|
AB_GLYCOPEPTIDES <- antibiotics %>%
|
|
filter(group %like% "glycopeptide") %>%
|
|
pull(ab)
|
|
AB_LIPOGLYCOPEPTIDES <- as.ab(c("DAL", "ORI", "TLV")) # dalba/orita/tela
|
|
AB_GLYCOPEPTIDES_EXCEPT_LIPO <- AB_GLYCOPEPTIDES[!AB_GLYCOPEPTIDES %in% AB_LIPOGLYCOPEPTIDES]
|
|
AB_LINCOSAMIDES <- antibiotics %>%
|
|
filter(atc_group2 %like% "lincosamide" | (group %like% "lincosamide" & is.na(atc_group2))) %>%
|
|
pull(ab)
|
|
AB_MACROLIDES <- antibiotics %>%
|
|
filter(atc_group2 %like% "macrolide" | (group %like% "macrolide" & is.na(atc_group2))) %>%
|
|
pull(ab)
|
|
AB_OXAZOLIDINONES <- antibiotics %>%
|
|
filter(group %like% "oxazolidinone") %>%
|
|
pull(ab)
|
|
AB_PENICILLINS <- antibiotics %>%
|
|
filter(group %like% "penicillin") %>%
|
|
pull(ab)
|
|
AB_POLYMYXINS <- antibiotics %>%
|
|
filter(group %like% "polymyxin") %>%
|
|
pull(ab)
|
|
AB_QUINOLONES <- antibiotics %>%
|
|
filter(group %like% "quinolone") %>%
|
|
pull(ab)
|
|
AB_STREPTOGRAMINS <- antibiotics %>%
|
|
filter(atc_group2 %like% "streptogramin") %>%
|
|
pull(ab)
|
|
AB_TETRACYCLINES <- antibiotics %>%
|
|
filter(group %like% "tetracycline") %>%
|
|
pull(ab)
|
|
AB_TETRACYCLINES_EXCEPT_TGC <- AB_TETRACYCLINES[AB_TETRACYCLINES != "TGC"]
|
|
AB_TRIMETHOPRIMS <- antibiotics %>%
|
|
filter(group %like% "trimethoprim") %>%
|
|
pull(ab)
|
|
AB_UREIDOPENICILLINS <- as.ab(c("PIP", "TZP", "AZL", "MEZ"))
|
|
AB_BETALACTAMS <- c(AB_PENICILLINS, AB_CEPHALOSPORINS, AB_CARBAPENEMS)
|
|
# this will be used for documentation:
|
|
DEFINED_AB_GROUPS <- ls(envir = globalenv())
|
|
DEFINED_AB_GROUPS <- DEFINED_AB_GROUPS[!DEFINED_AB_GROUPS %in% globalenv_before_ab]
|
|
create_AB_lookup <- function() {
|
|
AB_lookup <- AMR::antibiotics
|
|
AB_lookup$generalised_name <- generalise_antibiotic_name(AB_lookup$name)
|
|
AB_lookup$generalised_synonyms <- lapply(AB_lookup$synonyms, generalise_antibiotic_name)
|
|
AB_lookup$generalised_abbreviations <- lapply(AB_lookup$abbreviations, generalise_antibiotic_name)
|
|
AB_lookup$generalised_loinc <- lapply(AB_lookup$loinc, generalise_antibiotic_name)
|
|
AB_lookup$generalised_all <- unname(lapply(
|
|
as.list(as.data.frame(t(AB_lookup[,
|
|
c(
|
|
"ab", "atc", "cid", "name",
|
|
colnames(AB_lookup)[colnames(AB_lookup) %like% "generalised"]
|
|
),
|
|
drop = FALSE
|
|
]),
|
|
stringsAsFactors = FALSE
|
|
)),
|
|
function(x) {
|
|
x <- generalise_antibiotic_name(unname(unlist(x)))
|
|
x[x != ""]
|
|
}
|
|
))
|
|
AB_lookup[, colnames(AB_lookup)[colnames(AB_lookup) %like% "^generalised"]]
|
|
}
|
|
AB_LOOKUP <- create_AB_lookup()
|
|
|
|
# Export to package as internal data ----
|
|
usethis::ui_info(paste0("Updating internal package data"))
|
|
suppressMessages(usethis::use_data(EUCAST_RULES_DF,
|
|
TRANSLATIONS,
|
|
LANGUAGES_SUPPORTED_NAMES,
|
|
LANGUAGES_SUPPORTED,
|
|
MO_CONS,
|
|
MO_COPS,
|
|
MO_STREP_ABCG,
|
|
MO_FULLNAME_LOWER,
|
|
MO_PREVALENT_GENERA,
|
|
AB_LOOKUP,
|
|
AB_AMINOGLYCOSIDES,
|
|
AB_AMINOPENICILLINS,
|
|
AB_ANTIFUNGALS,
|
|
AB_ANTIMYCOBACTERIALS,
|
|
AB_CARBAPENEMS,
|
|
AB_CEPHALOSPORINS,
|
|
AB_CEPHALOSPORINS_1ST,
|
|
AB_CEPHALOSPORINS_2ND,
|
|
AB_CEPHALOSPORINS_3RD,
|
|
AB_CEPHALOSPORINS_4TH,
|
|
AB_CEPHALOSPORINS_5TH,
|
|
AB_CEPHALOSPORINS_EXCEPT_CAZ,
|
|
AB_FLUOROQUINOLONES,
|
|
AB_LIPOGLYCOPEPTIDES,
|
|
AB_GLYCOPEPTIDES,
|
|
AB_GLYCOPEPTIDES_EXCEPT_LIPO,
|
|
AB_LINCOSAMIDES,
|
|
AB_MACROLIDES,
|
|
AB_OXAZOLIDINONES,
|
|
AB_PENICILLINS,
|
|
AB_POLYMYXINS,
|
|
AB_QUINOLONES,
|
|
AB_STREPTOGRAMINS,
|
|
AB_TETRACYCLINES,
|
|
AB_TETRACYCLINES_EXCEPT_TGC,
|
|
AB_TRIMETHOPRIMS,
|
|
AB_UREIDOPENICILLINS,
|
|
AB_BETALACTAMS,
|
|
DEFINED_AB_GROUPS,
|
|
internal = TRUE,
|
|
overwrite = TRUE,
|
|
version = 2,
|
|
compress = "xz"
|
|
))
|
|
|
|
# Export data sets to the repository in different formats -----------------
|
|
|
|
for (pkg in c("haven", "openxlsx", "arrow")) {
|
|
if (!pkg %in% rownames(utils::installed.packages())) {
|
|
message("NOTE: package '", pkg, "' not installed! Ignoring export where this package is required.")
|
|
}
|
|
}
|
|
if ("digest" %in% rownames(utils::installed.packages())) {
|
|
md5 <- function(object) digest::digest(object, "md5")
|
|
} else {
|
|
# will write all files anyway, since MD5 hash cannot be determined
|
|
md5 <- function(object) "unknown-md5-hash"
|
|
}
|
|
|
|
write_md5 <- function(object) {
|
|
conn <- file(paste0("data-raw/", deparse(substitute(object)), ".md5"))
|
|
writeLines(md5(object), conn)
|
|
close(conn)
|
|
}
|
|
changed_md5 <- function(object) {
|
|
tryCatch(
|
|
{
|
|
conn <- file(paste0("data-raw/", deparse(substitute(object)), ".md5"))
|
|
compared <- md5(object) != readLines(con = conn)
|
|
close(conn)
|
|
compared
|
|
},
|
|
error = function(e) TRUE
|
|
)
|
|
}
|
|
|
|
# give official names to ABs and MOs
|
|
rsi <- rsi_translation %>%
|
|
mutate(mo_name = mo_name(mo, language = NULL, keep_synonyms = TRUE, info = FALSE), .after = mo) %>%
|
|
mutate(ab_name = ab_name(ab, language = NULL), .after = ab)
|
|
if (changed_md5(rsi)) {
|
|
usethis::ui_info(paste0("Saving {usethis::ui_value('rsi_translation')} to {usethis::ui_value('data-raw/')}"))
|
|
write_md5(rsi)
|
|
try(saveRDS(rsi, "data-raw/rsi_translation.rds", version = 2, compress = "xz"), silent = TRUE)
|
|
try(write.table(rsi, "data-raw/rsi_translation.txt", sep = "\t", na = "", row.names = FALSE), silent = TRUE)
|
|
try(haven::write_sas(rsi, "data-raw/rsi_translation.sas"), silent = TRUE)
|
|
try(haven::write_sav(rsi, "data-raw/rsi_translation.sav"), silent = TRUE)
|
|
try(haven::write_dta(rsi, "data-raw/rsi_translation.dta"), silent = TRUE)
|
|
try(openxlsx::write.xlsx(rsi, "data-raw/rsi_translation.xlsx"), silent = TRUE)
|
|
try(arrow::write_feather(rsi, "data-raw/rsi_translation.feather"), silent = TRUE)
|
|
try(arrow::write_parquet(rsi, "data-raw/rsi_translation.parquet"), silent = TRUE)
|
|
}
|
|
|
|
if (changed_md5(microorganisms)) {
|
|
usethis::ui_info(paste0("Saving {usethis::ui_value('microorganisms')} to {usethis::ui_value('data-raw/')}"))
|
|
write_md5(microorganisms)
|
|
try(saveRDS(microorganisms, "data-raw/microorganisms.rds", version = 2, compress = "xz"), silent = TRUE)
|
|
try(write.table(mo, "data-raw/microorganisms.txt", sep = "\t", na = "", row.names = FALSE), silent = TRUE)
|
|
max_50_snomed <- sapply(microorganisms$snomed, function(x) paste(x[seq_len(min(50, length(x), na.rm = TRUE))], collapse = " "))
|
|
mo <- microorganisms
|
|
mo$snomed <- max_50_snomed
|
|
mo <- dplyr::mutate_if(mo, ~ !is.numeric(.), as.character)
|
|
try(haven::write_sas(mo, "data-raw/microorganisms.sas"), silent = TRUE)
|
|
try(haven::write_sav(mo, "data-raw/microorganisms.sav"), silent = TRUE)
|
|
try(haven::write_dta(mo, "data-raw/microorganisms.dta"), silent = TRUE)
|
|
try(openxlsx::write.xlsx(mo, "data-raw/microorganisms.xlsx"), silent = TRUE)
|
|
try(arrow::write_feather(microorganisms, "data-raw/microorganisms.feather"), silent = TRUE)
|
|
try(arrow::write_parquet(microorganisms, "data-raw/microorganisms.parquet"), silent = TRUE)
|
|
}
|
|
|
|
ab <- dplyr::mutate_if(antibiotics, ~ !is.numeric(.), as.character)
|
|
if (changed_md5(ab)) {
|
|
usethis::ui_info(paste0("Saving {usethis::ui_value('antibiotics')} to {usethis::ui_value('data-raw/')}"))
|
|
write_md5(ab)
|
|
try(saveRDS(antibiotics, "data-raw/antibiotics.rds", version = 2, compress = "xz"), silent = TRUE)
|
|
try(write.table(antibiotics, "data-raw/antibiotics.txt", sep = "\t", na = "", row.names = FALSE), silent = TRUE)
|
|
try(haven::write_sas(ab, "data-raw/antibiotics.sas"), silent = TRUE)
|
|
try(haven::write_sav(ab, "data-raw/antibiotics.sav"), silent = TRUE)
|
|
try(haven::write_dta(ab, "data-raw/antibiotics.dta"), silent = TRUE)
|
|
try(openxlsx::write.xlsx(ab, "data-raw/antibiotics.xlsx"), silent = TRUE)
|
|
try(arrow::write_feather(antibiotics, "data-raw/antibiotics.feather"), silent = TRUE)
|
|
try(arrow::write_parquet(antibiotics, "data-raw/antibiotics.parquet"), silent = TRUE)
|
|
}
|
|
|
|
av <- dplyr::mutate_if(antivirals, ~ !is.numeric(.), as.character)
|
|
if (changed_md5(av)) {
|
|
usethis::ui_info(paste0("Saving {usethis::ui_value('antivirals')} to {usethis::ui_value('data-raw/')}"))
|
|
write_md5(av)
|
|
try(saveRDS(antivirals, "data-raw/antivirals.rds", version = 2, compress = "xz"), silent = TRUE)
|
|
try(write.table(av, "data-raw/antivirals.txt", sep = "\t", na = "", row.names = FALSE), silent = TRUE)
|
|
try(haven::write_sas(av, "data-raw/antivirals.sas"), silent = TRUE)
|
|
try(haven::write_sav(av, "data-raw/antivirals.sav"), silent = TRUE)
|
|
try(haven::write_dta(av, "data-raw/antivirals.dta"), silent = TRUE)
|
|
try(openxlsx::write.xlsx(av, "data-raw/antivirals.xlsx"), silent = TRUE)
|
|
try(arrow::write_feather(antivirals, "data-raw/antivirals.feather"), silent = TRUE)
|
|
try(arrow::write_parquet(antivirals, "data-raw/antivirals.parquet"), silent = TRUE)
|
|
}
|
|
|
|
# give official names to ABs and MOs
|
|
intrinsicR <- data.frame(
|
|
microorganism = mo_name(intrinsic_resistant$mo, language = NULL, keep_synonyms = TRUE, info = FALSE),
|
|
antibiotic = ab_name(intrinsic_resistant$ab, language = NULL),
|
|
stringsAsFactors = FALSE
|
|
)
|
|
if (changed_md5(intrinsicR)) {
|
|
usethis::ui_info(paste0("Saving {usethis::ui_value('intrinsic_resistant')} to {usethis::ui_value('data-raw/')}"))
|
|
write_md5(intrinsicR)
|
|
try(saveRDS(intrinsicR, "data-raw/intrinsic_resistant.rds", version = 2, compress = "xz"), silent = TRUE)
|
|
try(write.table(intrinsicR, "data-raw/intrinsic_resistant.txt", sep = "\t", na = "", row.names = FALSE), silent = TRUE)
|
|
try(haven::write_sas(intrinsicR, "data-raw/intrinsic_resistant.sas"), silent = TRUE)
|
|
try(haven::write_sav(intrinsicR, "data-raw/intrinsic_resistant.sav"), silent = TRUE)
|
|
try(haven::write_dta(intrinsicR, "data-raw/intrinsic_resistant.dta"), silent = TRUE)
|
|
try(openxlsx::write.xlsx(intrinsicR, "data-raw/intrinsic_resistant.xlsx"), silent = TRUE)
|
|
try(arrow::write_feather(intrinsicR, "data-raw/intrinsic_resistant.feather"), silent = TRUE)
|
|
try(arrow::write_parquet(intrinsicR, "data-raw/intrinsic_resistant.parquet"), silent = TRUE)
|
|
}
|
|
|
|
if (changed_md5(dosage)) {
|
|
usethis::ui_info(paste0("Saving {usethis::ui_value('dosage')} to {usethis::ui_value('data-raw/')}"))
|
|
write_md5(dosage)
|
|
try(saveRDS(dosage, "data-raw/dosage.rds", version = 2, compress = "xz"), silent = TRUE)
|
|
try(write.table(dosage, "data-raw/dosage.txt", sep = "\t", na = "", row.names = FALSE), silent = TRUE)
|
|
try(haven::write_sas(dosage, "data-raw/dosage.sas"), silent = TRUE)
|
|
try(haven::write_sav(dosage, "data-raw/dosage.sav"), silent = TRUE)
|
|
try(haven::write_dta(dosage, "data-raw/dosage.dta"), silent = TRUE)
|
|
try(openxlsx::write.xlsx(dosage, "data-raw/dosage.xlsx"), silent = TRUE)
|
|
try(arrow::write_feather(dosage, "data-raw/dosage.feather"), silent = TRUE)
|
|
try(arrow::write_parquet(dosage, "data-raw/dosage.parquet"), silent = TRUE)
|
|
}
|
|
|
|
suppressMessages(reset_AMR_locale())
|
|
|
|
# remove leftovers from global env
|
|
current_globalenv <- ls(envir = globalenv())
|
|
rm(list = current_globalenv[!current_globalenv %in% old_globalenv])
|
|
rm(current_globalenv)
|
|
|
|
devtools::load_all(quiet = TRUE)
|
|
suppressMessages(set_AMR_locale("English"))
|
|
|
|
# Update URLs -------------------------------------------------------------
|
|
usethis::ui_info("Checking URLs for redirects")
|
|
invisible(capture.output(urlchecker::url_update()))
|
|
|
|
|
|
# Document pkg ------------------------------------------------------------
|
|
if (interactive()) {
|
|
usethis::ui_info("Documenting package")
|
|
suppressMessages(devtools::document(quiet = TRUE))
|
|
}
|
|
|
|
|
|
# Style pkg ---------------------------------------------------------------
|
|
if (interactive()) {
|
|
# only when sourcing this file ourselves
|
|
usethis::ui_info("Styling package")
|
|
styler::style_pkg(
|
|
style = styler::tidyverse_style,
|
|
filetype = c("R", "Rmd")
|
|
)
|
|
}
|
|
|
|
|
|
# Finished ----------------------------------------------------------------
|
|
usethis::ui_done("All done")
|
|
suppressMessages(reset_AMR_locale())
|