### An [R package](https://www.r-project.org) to simplify the analysis and prediction of Antimicrobial Resistance (AMR) and work with antibiotic properties by using evidence-based methods.
This R package was created for academic research by PhD students of the Faculty of Medical Sciences of the [University of Groningen](https://www.rug.nl) and the Medical Microbiology & Infection Prevention (MMBI) department of the [University Medical Center Groningen (UMCG)](https://www.umcg.nl).
Matthijs S. Berends <ahref="https://orcid.org/0000-0001-7620-1800"><imgsrc="https://cran.r-project.org/web/orcid.svg"height="16px"></a><sup>1,2,a</sup>,
Christian F. Luz <ahref="https://orcid.org/0000-0001-5809-5995"><imgsrc="https://cran.r-project.org/web/orcid.svg"height="16px"></a><sup>1,a</sup>,
<sup>1</sup> Department of Medical Microbiology, University of Groningen, University Medical Center Groningen, Groningen, the Netherlands - [rug.nl](http://www.rug.nl) [umcg.nl](http://www.umcg.nl)<br>
<sup>2</sup> Certe Medical Diagnostics & Advice, Groningen, the Netherlands - [certe.nl](http://www.certe.nl)<br>
1. It **cleanses existing data**, by transforming it to reproducible and profound *classes*, making the most efficient use of R. These functions all use artificial intelligence to guess results that you would expect:
* Use `as.mo` to get an ID of a microorganism. The IDs are human readable for the trained eye - the ID of *Klebsiella pneumoniae* is "B_KLBSL_PNE" (B stands for Bacteria) and the ID of *S. aureus* is "B_STPHY_AUR". The function takes almost any text as input that looks like the name or code of a microorganism like "E. coli", "esco" and "esccol". Even `as.mo("MRSA")` will return the ID of *S. aureus*. Moreover, it can group all coagulase negative and positive *Staphylococci*, and can transform *Streptococci* into Lancefield groups. To find bacteria based on your input, it uses Artificial Intelligence to look up values in the included ITIS data, consisting of more than 18,000 microorganisms.
* Use `as.rsi` to transform values to valid antimicrobial results. It produces just S, I or R based on your input and warns about invalid values. Even values like "<=0.002; S" (combined MIC/RSI) will result in "S".
* Use `as.mic` to cleanse your MIC values. It produces a so-called factor (called *ordinal* in SPSS) with valid MIC values as levels. A value like "<=0.002; S" (combined MIC/RSI) will result in "<=0.002".
* Use `as.atc` to get the ATC code of an antibiotic as defined by the WHO. This package contains a database with most LIS codes, official names, DDDs and even trade names of antibiotics. For example, the values "Furabid", "Furadantin", "nitro" all return the ATC code of Nitrofurantoine.
* Use `first_isolate` to identify the first isolates of every patient [using guidelines from the CLSI](https://clsi.org/standards/products/microbiology/documents/m39/) (Clinical and Laboratory Standards Institute).
* You can also identify first *weighted* isolates of every patient, an adjusted version of the CLSI guideline. This takes into account key antibiotics of every strain and compares them.
* Use `MDRO` (abbreviation of Multi Drug Resistant Organisms) to check your isolates for exceptional resistance with country-specific guidelines or EUCAST rules. Currently, national guidelines for Germany and the Netherlands are supported.
* The data set `microorganisms` contains the complete taxonomic tree of more than 18,000 microorganisms (bacteria, fungi/yeasts and protozoa). Furthermore, the colloquial name and Gram stain are available, which enables resistance analysis of e.g. different antibiotics per Gram stain. The package also contains functions to look up values in this data set like `mo_genus`, `mo_family`, `mo_gramstain` or even `mo_phylum`. As they use `as.mo` internally, they also use artificial intelligence. For example, `mo_genus("MRSA")` and `mo_genus("S. aureus")` will both return `"Staphylococcus"`. They also come with support for German, Dutch, French, Italian, Spanish and Portuguese. These functions can be used to add new variables to your data.
* The data set `antibiotics` contains the ATC code, LIS codes, official name, trivial name and DDD of both oral and parenteral administration. It also contains a total of 298 trade names. Use functions like `ab_name` and `ab_tradenames` to look up values. The `ab_*` functions use `as.atc` internally so they support AI to guess your expected result. For example, `ab_name("Fluclox")`, `ab_name("Floxapen")` and `ab_name("J01CF05")` will all return `"Flucloxacillin"`. These functions can again be used to add new variables to your data.
* Calculate the resistance (and even co-resistance) of microbial isolates with the `portion_R`, `portion_IR`, `portion_I`, `portion_SI` and `portion_S` functions. Similarly, the *amount* of isolates can be determined with the `count_R`, `count_IR`, `count_I`, `count_SI` and `count_S` functions. All these functions can be used [with the `dplyr` package](https://dplyr.tidyverse.org/#usage) (e.g. in conjunction with [`summarise`](https://dplyr.tidyverse.org/reference/summarise.html))
This package contains the **complete microbial taxonomic data** (with all seven taxonomic ranks - from subkingdom to subspecies) from the publicly available Integrated Taxonomic Information System (ITIS, https://www.itis.gov).
All (sub)species from the taxonomic kingdoms Bacteria, Fungi and Protozoa are included in this package, as well as all previously accepted names known to ITIS. Furthermore, the responsible authors and year of publication are available. This allows users to use authoritative taxonomic information for their data analysis on any microorganism, not only human pathogens.
All stable versions of this package [are published on CRAN](http://cran.r-project.org/package=AMR), the official R network with a peer-reviewed submission process.
(Note: Downloads measured only by [cran.rstudio.com](https://cran.rstudio.com/package=AMR), this excludes e.g. the official [cran.r-project.org](https://cran.r-project.org/package=AMR))
This is the latest **development version**. Although it may contain bugfixes and even new functions compared to the latest released version on CRAN, it is also subject to change and may be unstable or behave unexpectedly. Always consider this a beta version. All below 'badges' should be green:
All functions checked on Linux and macOS | [](https://travis-ci.org/msberends/AMR) | Travis CI, GmbH [[ref 1]](https://travis-ci.org/msberends/AMR)
All functions checked on Windows | [](https://ci.appveyor.com/project/msberends/AMR) | Appveyor Systems Inc. [[ref 2]](https://ci.appveyor.com/project/msberends/AMR)
This package contains two new S3 classes: `mic` for MIC values (e.g. from Vitek or Phoenix) and `rsi` for antimicrobial drug interpretations (i.e. S, I and R). Both are actually ordered factors under the hood (an MIC of `2` being higher than `<=1` but lower than `>=32`, and for class `rsi` factors are ordered as `S < I < R`).
Both classes have extensions for existing generic functions like `print`, `summary` and `plot`.
The `septic_patients` data set comes with antimicrobial results of more than 40 different drugs. For example, columns `amox` and `cipr` contain results of amoxicillin and ciprofloxacin, respectively.
It also supports grouping variables. Let's say we want to compare resistance of drugs against Urine Tract Infections (UTI) between hospitals A to D (variable `hospital_id`):
Is there a significant difference between hospital A and D when it comes to Fosfomycin?
```r
check_A_and_D <-septic_patients%>%
filter(hospital_id %in% c("A", "D")) %>% # filter on only hospitals A and D
select(hospital_id, fosf) %>% # select the hospitals and fosfomycin
group_by(hospital_id) %>%
count_df(combine_IR = TRUE) %>% # count all isolates per group (hospital_id)
tidyr::spread(hospital_id, Value) %>% # transform output so A and D are columns
select(A, D) %>% # and select these only
as.matrix() # transform to good old matrix for fisher.test
check_A_and_D
# A D
# [1,] 24 33
# [2,] 25 77
```
Total sum is lower than 1,000 so we'd prefer a [Fisher's exact test](https://en.wikipedia.org/wiki/Fisher%27s_exact_test), not a [*G*-test](https://en.wikipedia.org/wiki/G-test) (or its formerly used equivalent, the famous [Chi<sup>2</sup> test](https://en.wikipedia.org/wiki/Chi-squared_test)):
Bacteria IDs can be retrieved with the `guess_mo` function. It uses any type of info about a microorganism as input. For example, all these will return value `B_STPHY_AUR`, the ID of *S. aureus*:
Base R lacks a simple function to create frequency tables. We created such a function that works with almost all data types: `freq` (or `frequency_tbl`). It can be used in two ways:
One of the most important features of this package is the complete microbial taxonomic database, supplied by ITIS (https://www.itis.gov). We created a function `as.mo` that transforms any user input value to a valid microbial ID by using AI (Artificial Intelligence) and based on the taxonomic tree of ITIS.
Using the `microbenchmark` package, we can review the calculation performance of this function.
```r
library(microbenchmark)
```
In the next test, we try to 'coerce' different input values for *Staphylococcus aureus*. The actual result is the same every time: it returns its MO code `B_STAPHY_AUR` (*B* stands for *Bacteria*, the taxonomic kingdom).
But the calculation time differs a lot. Here, the AI effect can be reviewed best:
```r
microbenchmark(A = as.mo("stau"),
B = as.mo("staaur"),
C = as.mo("S. aureus"),
D = as.mo("S. aureus"),
E = as.mo("STAAUR"),
F = as.mo("Staphylococcus aureus"),
times = 10,
unit = "ms")
# Unit: milliseconds
# expr min lq mean median uq max neval
# A 36.05088 36.14782 36.65635 36.24466 36.43075 39.78544 10
# B 16.43575 16.46885 16.67816 16.66053 16.84858 16.95299 10
# C 14.44150 14.52182 16.81197 14.59173 14.67854 36.75244 10
# D 14.49765 14.58153 16.71666 14.59414 14.61094 35.50731 10
# E 14.45212 14.75146 14.82033 14.85559 14.96433 15.03438 10
# F 10.69445 10.73852 10.80334 10.79596 10.86856 10.97465 10
The more an input value resembles a full name, the faster the result will be found. In the table above, all measurements are in milliseconds, tested on a quite regular Linux server from 2007 with 2 GB RAM. A value of 10.8 milliseconds means it will roughly determine 93 different (unique) input values per second. It case of 36.2 milliseconds, this is only 28 input values per second.
To improve speed, the `as.mo` function also takes into account the prevalence of human pathogenic microorganisms. The downside is of course that less prevalent microorganisms will be determined far less faster. See this example for the ID of *Burkholderia nodosa* (`B_BRKHL_NOD`):
```r
microbenchmark(B = as.mo("burnod"),
C = as.mo("B. nodosa"),
D = as.mo("B. nodosa"),
E = as.mo("BURNOD"),
F = as.mo("Burkholderia nodosa"),
times = 10,
unit = "ms")
# Unit: milliseconds
# expr min lq mean median uq max neval
# B 175.9446 176.80440 179.18240 177.00131 177.62021 198.86286 10
# C 88.1902 88.57705 89.08851 88.84293 89.15498 91.76621 10
# D 110.2641 110.67497 113.66290 111.20534 111.80744 134.44699 10
# E 175.0728 177.04235 207.83542 190.38109 200.33448 388.12177 10
# F 45.0778 45.31617 52.72430 45.62962 67.85262 70.42250 10
```
(Note: `A` is missing here, because `as.mo("buno")` returns `F_BUELL_NOT`: the ID of the fungus *Buellia notabilis*)
That takes up to 12 times as much time! A value of 190.4 milliseconds means it can only determine ~5 different input values per second. We can conclude that looking up arbitrary codes of less prevalent microorganisms is the worst way to go, in terms of calculation performance.
To relieve this pitfall and further improve performance, two important calculations take almost no time at all: **repetive results** and **already precalculated results**.
Let's set up 25,000 entries of `"Staphylococcus aureus"` and check its speed:
So transforming 25,000 times (!) `"Staphylococcus aureus"` only takes 4 ms (0.004 seconds) more than transforming it once. You only lose time on your unique input values.
What about precalculated results? This package also contains helper functions for specific microbial properties, for example `mo_fullname`. It returns the full microbial name (genus, species and possibly subspecies) and uses `as.mo` internally. If the input is however an already precalculated result, it almost doesn't take any time at all (see 'C' below):
```r
microbenchmark(A = mo_fullname("B_STPHY_AUR"),
B = mo_fullname("S. aureus"),
C = mo_fullname("Staphylococcus aureus"),
times = 10,
unit = "ms")
# Unit: milliseconds
# expr min lq mean median uq max neval
# A 13.548652 13.74588 13.8052969 13.813594 13.881165 14.090969 10
# B 15.079781 15.16785 15.3835842 15.374477 15.395115 16.072995 10
So going from `mo_fullname("Staphylococcus aureus")` to `"Staphylococcus aureus"` takes 0.0002 seconds - it doesn't even start calculating *if the result would be the same as the expected resulting value*. That goes for all helper functions:
```r
microbenchmark(A = mo_species("aureus"),
B = mo_genus("Staphylococcus"),
C = mo_fullname("Staphylococcus aureus"),
D = mo_family("Staphylococcaceae"),
E = mo_order("Bacillales"),
F = mo_class("Bacilli"),
G = mo_phylum("Firmicutes"),
H = mo_subkingdom("Posibacteria"),
times = 10,
unit = "ms")
# Unit: milliseconds
# expr min lq mean median uq max neval
# A 0.145270 0.158750 0.1908419 0.1693655 0.218255 0.300528 10
# B 0.182985 0.184522 0.2025408 0.1970235 0.209944 0.243328 10
# C 0.176280 0.201632 0.2618147 0.2303025 0.339499 0.388249 10
# D 0.136890 0.139054 0.1552231 0.1518010 0.168738 0.193042 10
# E 0.100921 0.116496 0.1321823 0.1222930 0.129976 0.230477 10
# F 0.103017 0.110281 0.1214480 0.1199880 0.124319 0.147506 10
# G 0.099246 0.110280 0.1195553 0.1188705 0.125436 0.149741 10
# H 0.114331 0.117264 0.1249819 0.1220830 0.129557 0.143385 10
```
Of course, when running `mo_phylum("Firmicutes")` the function has zero knowledge about the actual microorganism, namely *S. aureus*. But since the result would be `"Firmicutes"` too, there is no point in calculating the result. And since this package 'knows' all phyla of all known microorganisms (according to ITIS), it can just return the initial value immediately.
This R package is licensed under the [GNU General Public License (GPL) v2.0](https://github.com/msberends/AMR/blob/master/LICENSE). In a nutshell, this means that this package:
