mirror of
https://github.com/msberends/AMR.git
synced 2024-12-25 18:46:11 +01:00
185 lines
7.1 KiB
R
185 lines
7.1 KiB
R
# ==================================================================== #
|
|
# TITLE #
|
|
# AMR: An R Package for Working with Antimicrobial Resistance Data #
|
|
# #
|
|
# SOURCE #
|
|
# https://github.com/msberends/AMR #
|
|
# #
|
|
# CITE AS #
|
|
# Berends MS, Luz CF, Friedrich AW, Sinha BNM, Albers CJ, Glasner C #
|
|
# (2022). AMR: An R Package for Working with Antimicrobial Resistance #
|
|
# Data. Journal of Statistical Software, 104(3), 1-31. #
|
|
# doi:10.18637/jss.v104.i03 #
|
|
# #
|
|
# Developed at the University of Groningen and the University Medical #
|
|
# Center Groningen in The Netherlands, in collaboration with many #
|
|
# colleagues from around the world, see our website. #
|
|
# #
|
|
# 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/ #
|
|
# ==================================================================== #
|
|
|
|
library(dplyr)
|
|
library(tidyr)
|
|
library(rvest)
|
|
|
|
# get all data from the WHOCC website
|
|
get_atc_table <- function(atc_group) {
|
|
# give as input J0XXX, like atc_group = "J05AB"
|
|
downloaded <- read_html(paste0("https://www.whocc.no/atc_ddd_index/?code=", atc_group, "&showdescription=no"))
|
|
table_title <- downloaded %>%
|
|
html_nodes(paste0('a[href^="./?code=', atc_group, '&"]')) %>%
|
|
html_text()
|
|
table_title <- table_title[tolower(table_title) != "show text from guidelines"][1]
|
|
table_content <- downloaded %>%
|
|
html_nodes("table") %>%
|
|
html_table(header = TRUE) %>%
|
|
# returns list, so make data.frame out of it
|
|
as.data.frame(stringsAsFactors = FALSE) %>%
|
|
# select right columns
|
|
select(atc = ATC.code, name = Name, ddd = DDD, unit = U, ddd_type = Adm.R) %>%
|
|
# fill empty rows
|
|
mutate(atc = ifelse(atc == "", lag(atc), atc), name = ifelse(name == "", lag(name), name)) %>%
|
|
pivot_wider(names_from = ddd_type, values_from = c(ddd, unit)) %>%
|
|
mutate(atc_group = table_title)
|
|
if (!"ddd_O" %in% colnames(table_content)) {
|
|
table_content <- table_content %>% mutate(ddd_O = NA_real_, unit_O = NA_character_)
|
|
}
|
|
if (!"ddd_P" %in% colnames(table_content)) {
|
|
table_content <- table_content %>% mutate(ddd_P = NA_real_, unit_P = NA_character_)
|
|
}
|
|
table_content %>% select(atc, name, atc_group,
|
|
oral_ddd = ddd_O, oral_units = unit_O,
|
|
iv_ddd = ddd_P, iv_units = unit_P
|
|
)
|
|
}
|
|
|
|
# these are the relevant groups for input: https://www.whocc.no/atc_ddd_index/?code=J05A (J05 only contains J05A)
|
|
atc_groups <- c("J05AA", "J05AB", "J05AC", "J05AD", "J05AE", "J05AF", "J05AG", "J05AH", "J05AJ", "J05AP", "J05AR", "J05AX")
|
|
|
|
# get the first
|
|
antivirals <- get_atc_table(atc_groups[1])
|
|
# bind all others to it
|
|
for (i in 2:length(atc_groups)) {
|
|
message(atc_groups[i], "...")
|
|
antivirals <- rbind(antivirals, get_atc_table(atc_groups[i]))
|
|
}
|
|
|
|
# arrange on name, untibble it
|
|
antivirals <- antivirals %>%
|
|
arrange(name) %>%
|
|
as.data.frame(stringsAsFactors = FALSE)
|
|
|
|
# add PubChem Compound ID (cid) and their trade names
|
|
# see `data-raw/reproduction_of_antibiotics` for get_CID() and get_synonyms()
|
|
CIDs <- get_CID(antivirals$name)
|
|
# these could not be found:
|
|
antivirals[is.na(CIDs), ] %>% View()
|
|
# get brand names from PubChem
|
|
synonyms <- get_synonyms(CIDs)
|
|
synonyms <- lapply(
|
|
synonyms,
|
|
function(x) {
|
|
if (length(x) == 0 | all(is.na(x))) {
|
|
""
|
|
} else {
|
|
x
|
|
}
|
|
}
|
|
)
|
|
|
|
antivirals <- antivirals %>%
|
|
transmute(atc,
|
|
cid = as.double(CIDs),
|
|
name,
|
|
atc_group,
|
|
synonyms = unname(synonyms),
|
|
oral_ddd,
|
|
oral_units,
|
|
iv_ddd,
|
|
iv_units
|
|
) %>%
|
|
AMR:::dataset_UTF8_to_ASCII()
|
|
|
|
av_codes <- tibble(name = antivirals$name %>%
|
|
strsplit("(, | and )") %>%
|
|
unlist() %>%
|
|
unique() %>%
|
|
sort()) %>%
|
|
mutate(av_1st = toupper(abbreviate(name, minlength = 3, use.classes = FALSE))) %>%
|
|
filter(!name %in% c("acid", "dipivoxil", "disoproxil", "marboxil", "alafenamide"))
|
|
|
|
replace_with_av_code <- function(name) {
|
|
unname(av_codes$av_1st[match(name, av_codes$name)])
|
|
}
|
|
|
|
names_codes <- antivirals %>%
|
|
separate(name,
|
|
into = paste0("name", c(1:7)),
|
|
sep = "(, | and )",
|
|
remove = FALSE,
|
|
fill = "right") %>%
|
|
# remove empty columns
|
|
select(!where(function(x) all(is.na(x)))) %>%
|
|
mutate_at(vars(matches("name[1-9]")), replace_with_av_code) %>%
|
|
unite(av, matches("name[1-9]"), sep = "+", na.rm = TRUE) %>%
|
|
mutate(name = gsub("(, | and )", "/", name))
|
|
substr(names_codes$name, 1, 1) <- toupper(substr(names_codes$name, 1, 1))
|
|
|
|
antivirals <- bind_cols(
|
|
names_codes %>% select(av, name),
|
|
antivirals %>% select(-name)
|
|
)
|
|
class(antivirals$av) <- c("av", "character")
|
|
antivirals <- antivirals %>% AMR:::dataset_UTF8_to_ASCII()
|
|
|
|
# add loinc, see 'data-raw/loinc.R'
|
|
loinc_df <- read.csv("data-raw/Loinc.csv",
|
|
row.names = NULL,
|
|
stringsAsFactors = FALSE)
|
|
|
|
loinc_df <- loinc_df %>% filter(CLASS == "DRUG/TOX")
|
|
av_names <- antivirals %>%
|
|
pull(name) %>%
|
|
paste0(collapse = "|") %>%
|
|
paste0("(", ., ")")
|
|
|
|
antivirals$loinc <- as.list(rep(NA_character_, nrow(antivirals)))
|
|
for (i in seq_len(nrow(antivirals))) {
|
|
message(i)
|
|
loinc_ab <- loinc_df %>%
|
|
filter(COMPONENT %like% paste0("^", antivirals$name[i])) %>%
|
|
pull(LOINC_NUM)
|
|
if (length(loinc_ab) > 0) {
|
|
antivirals$loinc[i] <- list(loinc_ab)
|
|
}
|
|
}
|
|
# sort and fix for empty values
|
|
for (i in 1:nrow(antivirals)) {
|
|
loinc <- as.character(sort(unique(tolower(antivirals[i, "loinc", drop = TRUE][[1]]))))
|
|
antivirals[i, "loinc"][[1]] <- ifelse(length(loinc[!loinc == ""]) == 0, list(""), list(loinc))
|
|
}
|
|
|
|
# de-duplicate synonyms
|
|
for (i in 1:nrow(antivirals)) {
|
|
syn <- as.character(sort(unique(tolower(antivirals[i, "synonyms", drop = TRUE][[1]]))))
|
|
syn <- syn[!syn %in% tolower(antivirals[i, "name", drop = TRUE])]
|
|
antivirals[i, "synonyms"][[1]] <- ifelse(length(syn[!syn == ""]) == 0, list(""), list(syn))
|
|
}
|
|
|
|
antivirals <- antivirals %>% AMR:::dataset_UTF8_to_ASCII()
|
|
|
|
# check it
|
|
antivirals
|
|
|
|
# save it
|
|
usethis::use_data(antivirals, overwrite = TRUE, internal = FALSE, compress = "xz", version = 2)
|