* Much increased algorithms to translate user input to valid taxonomy, e.g. by using [recent scientific work](https://doi.org/10.1099/mic.0.001269) about per-species human pathogenicity
#### Interpretation of MIC and disk diffusion values
EUCAST 2022 and CLSI 2022 guidelines have been added for `as.rsi()`. EUCAST 2022 (v12.0) is now the new default guideline for all MIC and disks diffusion interpretations, and for `eucast_rules()` to apply EUCAST Expert Rules. The default guideline (EUCAST) can now be changed with the new `AMR_guideline` option, such as: `options(AMR_guideline = "CLSI 2020")`.
Interpretation guidelines older than 10 years were removed, the oldest now included guidelines of EUCAST and CLSI are from 2013.
#### Supported languages
We added support for the following languages: Chinese, Greek, Japanese, Polish, Turkish and Ukrainian. All antibiotic names are now available in these languages, and the AMR package will automatically determine a supported language based on the user system language.
We are very grateful for the valuable input by our colleagues from other countries. The `AMR` package is now available in 16 languages and according to download stats used in almost all countries in the world!
The `microorganisms` no longer relies on the Catalogue of Life, but on the List of Prokaryotic names with Standing in Nomenclature (LPSN) and is supplemented with the 'backbone taxonomy' from the Global Biodiversity Information Facility (GBIF). The structure of this data set has changed to include separate LPSN and GBIF identifiers. Almost all previous MO codes were retained. It contains over 1,400 taxonomic names from 2022.
We previously relied on our own experience to categorise species into pathogenic groups, but we were very happy to encounter the very recent work of Bartlett *et al.* (2022, DOI [10.1099/mic.0.001269](https://doi.org/10.1099/mic.0.001269)) who extensively studied medical-scientific literature to categorise all bacterial species into groups. See `mo_matching_score()` on how their work was incorporated into the `prevalence` column of the `microorganisms` data set. Using their results, the `as.mo()` and all `mo_*()` functions are now much better capable of converting user input to valid taxonomic records.
* Updated full microbiological taxonomy according to the latest daily LPSN data set (December 2022) and latest yearly GBIF taxonomy backbone (November 2022)
* Support for all 1,515 city-like serovars of *Salmonella*, such as *Salmonella* Goldcoast. Formally, these are serovars belonging to the *S. enterica* species, but they are reported with only the name of the genus and the city. For this reason, the serovars are in the `subspecies` column of the `microorganisms` data set and "enterica" is in the `species` column, but the full name does not contain the species name (*enterica*).
* All new algorithm for `as.mo()` (and thus all `mo_*()` functions) while still following our original set-up as described in our recently published JSS paper (DOI [10.18637/jss.v104.i03](https://doi.org/10.18637/jss.v104.i03)).
* A new argument `keep_synonyms` allows to *not* correct for updated taxonomy, in favour of the now deleted argument `allow_uncertain`
* It has increased tremendously in speed and returns generally more consequent results
* Sequential coercion is now extremely fast as results are stored to the package environment, although coercion of unknown values must be run once per session. Previous results can be reset/removed with the new `mo_reset_session()` function.
* The MO matching score algorithm (`mo_matching_score()`) now counts deletions and substitutions as 2 instead of 1, which impacts the outcome of `as.mo()` and any `mo_*()` function
* **Removed all species of the taxonomic kingdom Chromista** from the package. This was done for multiple reasons:
* CRAN allows packages to be around 5 MB maximum, some packages are exempted but this package is not one of them
* Chromista are not relevant when it comes to antimicrobial resistance, thus lacking the primary scope of this package
* Chromista are almost never clinically relevant, thus lacking the secondary scope of this package
* The `microorganisms.old` data set was removed, and all previously accepted names are now included in the `microorganisms` data set. A new column `status` contains `"accepted"` for currently accepted names and `"synonym"` for taxonomic synonyms; currently invalid names. All previously accepted names now have a microorganisms ID and - if available - an LPSN, GBIF and SNOMED CT identifier.
#### Antibiotic agents and selectors
The new function `add_custom_antimicrobials()` allows users to add custom antimicrobial codes and names to the `AMR` package.
* Antibiotic code "CEI" for ceftolozane/tazobactam has been replaced with "CZT" to comply with EARS-Net and WHONET 2022. The old code will still work in all cases when using `as.ab()` or any of the `ab_*()` functions.
* Support for antimicrobial interpretation of anaerobic bacteria, by adding a 'placeholder' code `B_ANAER` to the `microorganisms` data set and adding the breakpoints of anaerobics to the `rsi_interpretation` data set, which is used by `as.rsi()` for interpretion of MIC and disk diffusion values
Also, we added support for using antibiotic selectors in scoped `dplyr` verbs (with or without using `vars()`), such as in: `... %>% summarise_at(aminoglycosides(), resistance)`, please see `resistance()` for examples.
#### Antiviral agents
We now added extensive support for antiviral agents! For the first time, the `AMR` package has extensive support for antiviral drugs and to work with their names, codes and other data in any way.
* The `antivirals` data set has been extended with 18 new drugs (also from the [new J05AJ ATC group](https://www.whocc.no/atc_ddd_index/?code=J05AJ&showdescription=no)) and now also contains antiviral identifiers and LOINC codes
* A new data type `av` (*antivirals*) has been added, which is functionally similar to `ab` for antibiotics
* Functions `as.av()`, `av_name()`, `av_atc()`, `av_synonyms()`, `av_from_text()` have all been added as siblings to their `ab_*()` equivalents
#### Other new functions
* Function `rsi_confidence_interval()` to add confidence intervals in AMR calculation. This is now also included in `rsi_df()` and `proportion_df()`.
* Function `mean_amr_distance()` to calculate the mean AMR distance. The mean AMR distance is a normalised numeric value to compare AMR test results and can help to identify similar isolates, without comparing antibiograms by hand.
* Function `rsi_interpretation_history()` to view the history of previous runs of `as.rsi()`. This returns a 'logbook' with the selected guideline, reference table and specific interpretation of each row in a data set on which `as.rsi()` was run.
* Function `mo_current()` to get the currently valid taxonomic name of a microorganism
* Function `add_custom_antimicrobials()` to add custom antimicrobial codes and names to the `AMR` package
* Argument `combine_IR` has been removed from this package (affecting functions `count_df()`, `proportion_df()`, and `rsi_df()` and some plotting functions), since it was replaced with `combine_SI` three years ago
* Using `units` in `ab_ddd(..., units = "...")` had been deprecated for some time and is now not supported anymore. Use `ab_ddd_units()` instead.
* Support for `data.frame`-enhancing R packages, more specifically: `data.table::data.table`, `janitor::tabyl`, `tibble::tibble`, and `tsibble::tsibble`. AMR package functions that have a data set as output (such as `rsi_df()` and `bug_drug_combinations()`), will now return the same data type as the input.
* All data sets in this package are now a `tibble`, instead of base R `data.frame`s. Older R versions are still supported, even if they do not support `tibble`s.
* Our data sets are now also continually exported to **Apache Feather and Apache Parquet formats**. You can find more info [in this article on our website](https://msberends.github.io/AMR/articles/datasets.html).
* Interpretation from MIC and disk zones is now more informative about availability of breakpoints and more robust
* Removed the `as.integer()` method for MIC values, since MIC are not integer values and running `table()` on MIC values consequently failed for not being able to retrieve the level position (as that's how normally `as.integer()` on `factor`s work)
* Fixed determination of Gram stains (`mo_gramstain()`), since the taxonomic phyla Actinobacteria, Chloroflexi, Firmicutes, and Tenericutes have been renamed to respectively Actinomycetota, Chloroflexota, Bacillota, and Mycoplasmatota in 2021
*`droplevels()` on MIC will now return a common `factor` at default and will lose the `mic` class. Use `droplevels(..., as.mic = TRUE)` to keep the `mic` class.
* Fixes for reading in text files using `set_mo_source()`, which now also allows the source file to contain valid taxonomic names instead of only valid microorganism ID of this package
* Using any `random_*()` function (such as `random_mic()`) is now possible by directly calling the package without loading it first: `AMR::random_mic(10)`
* Extended support for the `vctrs` package, used internally by the tidyverse. This allows to change values of class `mic`, `disk`, `rsi`, `mo` and `ab` in tibbles, and to use antibiotic selectors for selecting/filtering, e.g. `df[carbapenems() == "R", ]`
* New website to make use of the new Bootstrap 5 and pkgdown 2.0. The website now contains results for all examples and will be automatically regenerated with every change to our repository, using GitHub Actions
