AMR/data-raw/read_EUCAST.R

# ==================================================================== #
# TITLE:                                                               #
# AMR: An R Package for Working with Antimicrobial Resistance Data     #
#                                                                      #
# SOURCE CODE:                                                         #
# https://github.com/msberends/AMR                                     #
#                                                                      #
# PLEASE CITE THIS SOFTWARE AS:                                        #
# Berends MS, Luz CF, Friedrich AW, et al. (2022).                     #
# AMR: An R Package for Working with Antimicrobial Resistance Data.    #
# Journal of Statistical Software, 104(3), 1-31.                       #
# https://doi.org/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(openxlsx)
library(dplyr)
library(tidyr)
library(cleaner)
library(AMR)

# USE THIS FUNCTION TO READ THE EUCAST EXCEL FILE THAT CONTAINS THE BREAKPOINT TABLES

read_EUCAST <- function(sheet, file, guideline_name) {
  message("\nGetting sheet: ", sheet)
  sheet.bak <- sheet

  uncertainties <- NULL
  add_uncertainties <- function(old, new) {
    if (is.null(old)) {
      new
    } else {
      bind_rows(old, new)
    }
  }

  raw_data <- read.xlsx(
    xlsxFile = file,
    sheet = sheet,
    colNames = FALSE,
    skipEmptyRows = FALSE,
    skipEmptyCols = FALSE,
    fillMergedCells = TRUE,
    na.strings = c("", "-", "NA", "IE", "IP")
  )
  probable_rows <- suppressWarnings(raw_data %>% mutate_all(as.double) %>% summarise_all(~ sum(!is.na(.))) %>% unlist() %>% max())
  if (probable_rows == 0) {
    message("NO ROWS FOUND")
    message("------------------------")
    return(NULL)
  }

  # in the info header in the Excel file, EUCAST mentions which genera are targeted
  if (sheet %like% "anaerob.*Gram.*posi") {
    sheet <- paste0(
      c(
        "Actinomyces", "Bifidobacterium", "Clostridioides",
        "Clostridium", "Cutibacterium", "Eggerthella",
        "Eubacterium", "Lactobacillus", "Propionibacterium",
        "Staphylococcus saccharolyticus"
      ),
      collapse = "_"
    )
  } else if (sheet %like% "anaerob.*Gram.*nega") {
    sheet <- paste0(
      c(
        "Bacteroides",
        "Bilophila",
        "Fusobacterium",
        "Mobiluncus",
        "Parabacteroides",
        "Porphyromonas",
        "Prevotella"
      ),
      collapse = "_"
    )
  } else if (sheet == "Streptococcus A,B,C,G") {
    sheet <- paste0(
      microorganisms %>%
        filter(genus == "Streptococcus") %>%
        mutate(lancefield = mo_name(mo, Lancefield = TRUE)) %>%
        filter(lancefield %like% "^Streptococcus group") %>%
        pull(fullname),
      collapse = "_"
    )
  } else if (sheet %like% "PK.*PD") {
    sheet <- "UNKNOWN"
  }
  mo_sheet <- paste0(suppressMessages(as.mo(unlist(strsplit(sheet, "_")))), collapse = "|")
  if (!is.null(mo_uncertainties())) uncertainties <- add_uncertainties(uncertainties, mo_uncertainties())

  set_columns_names <- function(x, cols) {
    colnames(x) <- cols[1:length(colnames(x))]
    x
  }

  get_mo <- function(x) {
    for (i in seq_len(length(x))) {
      y <- trimws2(unlist(strsplit(x[i], "(,|and)")))
      y <- trimws2(gsub("[(].*[)]", "", y))
      y <- suppressWarnings(suppressMessages(as.mo(y)))
      if (!is.null(mo_uncertainties())) uncertainties <<- add_uncertainties(uncertainties, mo_uncertainties())
      y <- y[!is.na(y) & y != "UNKNOWN"]
      x[i] <- paste(y, collapse = "|")
    }
    x
  }

  MICs_with_trailing_superscript <- c(
    seq(from = 0.0011, to = 0.0019, by = 0.0001),
    seq(from = 0.031, to = 0.039, by = 0.001),
    seq(from = 0.061, to = 0.069, by = 0.001),
    seq(from = 0.1251, to = 0.1259, by = 0.0001),
    seq(from = 0.251, to = 0.259, by = 0.001),
    seq(from = 0.51, to = 0.59, by = 0.01),
    seq(from = 11, to = 19, by = 1),
    seq(from = 161, to = 169, by = 01),
    seq(from = 21, to = 29, by = 1),
    seq(from = 321, to = 329, by = 1),
    seq(from = 41, to = 49, by = 1),
    seq(from = 81, to = 89, by = 1)
  )
  has_superscript <- function(x) {
    # because due to floating point error, 0.1252 is not in:
    # seq(from = 0.1251, to = 0.1259, by = 0.0001)
    sapply(x, function(x) any(near(x, MICs_with_trailing_superscript)))
  }

  has_zone_diameters <- rep(any(unlist(raw_data) %like% "zone diameter"), nrow(raw_data))

  cleaned <- raw_data %>%
    as_tibble() %>%
    set_columns_names(LETTERS) %>%
    transmute(
      drug = A,
      MIC_S = B,
      MIC_R = C,
      disk_dose = ifelse(has_zone_diameters, E, NA_character_),
      disk_S = ifelse(has_zone_diameters, `F`, NA_character_),
      disk_R = ifelse(has_zone_diameters, G, NA_character_)
    ) %>%
    filter(
      !is.na(drug),
      !(is.na(MIC_S) & is.na(MIC_R) & is.na(disk_S) & is.na(disk_R)),
      MIC_S %unlike% "(MIC|S ≤|note)",
      MIC_S %unlike% "^[-]",
      drug != MIC_S,
    ) %>%
    mutate(
      administration = case_when(
        drug %like% "[( ]oral" ~ "oral",
        drug %like% "[( ]iv" ~ "iv",
        TRUE ~ NA_character_
      ),
      uti = ifelse(drug %like% "(UTI|urinary|urine)", TRUE, FALSE),
      systemic = ifelse(drug %like% "(systemic|septic)", TRUE, FALSE),
      mo = ifelse(drug %like% "([.]|spp)", get_mo(drug), mo_sheet)
    ) %>%
    # clean disk doses
    mutate(disk_dose = clean_character(disk_dose, remove = "[^0-9.-]")) %>%
    # clean MIC and disk values
    mutate(
      MIC_S = gsub(".,.", "", MIC_S), # remove superscript notes with comma, like 0.5^2,3
      MIC_R = gsub(".,.", "", MIC_R),
      disk_S = gsub(".,.", "", disk_S),
      disk_R = gsub(".,.", "", disk_R),
      MIC_S = clean_double(MIC_S), # make them valid numeric values
      MIC_R = clean_double(MIC_R),
      disk_S = clean_integer(disk_S),
      disk_R = clean_integer(disk_R),
      # invalid MIC values have a superscript text, delete those
      MIC_S = ifelse(has_superscript(MIC_S),
        substr(MIC_S, 1, nchar(MIC_S) - 1),
        MIC_S
      ),
      MIC_R = ifelse(has_superscript(MIC_R),
        substr(MIC_R, 1, nchar(MIC_R) - 1),
        MIC_R
      ),
      # and some are just awful
      MIC_S = ifelse(MIC_S == 43.4, 4, MIC_S),
      MIC_R = ifelse(MIC_R == 43.4, 4, MIC_R),
    ) %>%
    # clean drug names
    mutate(
      drug = gsub(" ?[(, ].*$", "", drug),
      drug = gsub("[1-9]+$", "", drug),
      ab = as.ab(drug)
    ) %>%
    select(ab, mo, everything(), -drug) %>%
    as.data.frame(stringsAsFactors = FALSE)

  # new row for every different MO mentioned
  for (i in 1:nrow(cleaned)) {
    mo <- cleaned[i, "mo", drop = TRUE]
    if (grepl(pattern = "|", mo, fixed = TRUE)) {
      mo_vect <- unlist(strsplit(mo, "|", fixed = TRUE))
      cleaned[i, "mo"] <- mo_vect[1]
      for (j in seq_len(length(mo_vect))) {
        cleaned <- bind_rows(cleaned, cleaned[i, , drop = FALSE])
        cleaned[nrow(cleaned), "mo"] <- mo_vect[j]
      }
    }
  }

  cleaned <- cleaned %>%
    distinct(ab, mo, administration, uti, systemic, .keep_all = TRUE) %>%
    arrange(ab, mo) %>%
    mutate_at(c("MIC_S", "MIC_R", "disk_S", "disk_R"), as.double) %>%
    pivot_longer(c("MIC_S", "MIC_R", "disk_S", "disk_R"), "type") %>%
    mutate(
      method = ifelse(type %like% "MIC", "MIC", "DISK"),
      type = gsub("^.*_", "breakpoint_", type)
    ) %>%
    pivot_wider(names_from = type, values_from = value) %>%
    mutate(
      guideline = guideline_name,
      disk_dose = ifelse(method == "DISK", disk_dose, NA_character_),
      mo = ifelse(mo == "", mo_sheet, mo)
    ) %>%
    filter(!(is.na(breakpoint_S) & is.na(breakpoint_R))) %>%
    # comply with clinical_breakpoints for now
    transmute(guideline,
      method,
      site = case_when(
        uti ~ "UTI",
        systemic ~ "Systemic",
        TRUE ~ administration
      ),
      mo, ab,
      ref_tbl = sheet.bak,
      disk_dose = ifelse(!is.na(disk_dose), paste0(disk_dose, "ug"), NA_character_),
      breakpoint_S,
      breakpoint_R
    ) %>%
    as.data.frame(stringsAsFactors = FALSE)

  if (!is.null(uncertainties)) {
    print(uncertainties %>% distinct(input, mo, .keep_all = TRUE))
  }

  message("Estimated: ", probable_rows, ", gained: ", cleaned %>% count(ab) %>% nrow())
  message("------------------------")
  cleaned
}


# Actual import -----------------------------------------------------------

file <- "data-raw/v_11.0_Breakpoint_Tables.xlsx"
sheets <- readxl::excel_sheets(file)
guideline_name <- "EUCAST 2021"

sheets_to_analyse <- sheets[!sheets %in% c("Content", "Changes", "Notes", "Guidance", "Dosages", "Technical uncertainty", "Topical agents")]

# takes the longest time:
new_EUCAST <- read_EUCAST(
  sheet = sheets_to_analyse[1],
  file = file,
  guideline_name = guideline_name
)
for (i in 2:length(sheets_to_analyse)) {
  tryCatch(
    new_EUCAST <<- bind_rows(
      new_EUCAST,
      read_EUCAST(
        sheet = sheets_to_analyse[i],
        file = file,
        guideline_name = guideline_name
      )
    ),
    error = function(e) message(e$message)
  )
}

# 2021-07-12 fix for Morganellaceae (check other lines too next time)
morg <- clinical_breakpoints %>%
  as_tibble() %>%
  filter(
    ab == "IPM",
    guideline == "EUCAST 2021",
    mo == as.mo("Enterobacterales")
  ) %>%
  mutate(mo = as.mo("Morganellaceae"))
morg[which(morg$method == "MIC"), "breakpoint_S"] <- 0.001
morg[which(morg$method == "MIC"), "breakpoint_R"] <- 4
morg[which(morg$method == "DISK"), "breakpoint_S"] <- 50
morg[which(morg$method == "DISK"), "breakpoint_R"] <- 19

clinical_breakpoints <- clinical_breakpoints %>%
  bind_rows(morg) %>%
  bind_rows(morg %>%
    mutate(guideline = "EUCAST 2020")) %>%
  arrange(desc(guideline), ab, mo, method)