R/geo_load_esets.R

In miracum/clearly-sigident.preproc: Sigident Preprocessing

#' @title geo_load_eset
#'
#' @description Helper function to load expression sets from GEO.
#'
#' @param name A character string. The name of the GEO study.
#' @param datadir A character string. The path to the directory to store
#'   the downloaded GEO datasets.
#' @param targetcolname A character string. The name of the column in the
#'   GEO expression set's pheno data, holding the variable of interest
#'   (mapping information).
#' @param targetlevelname A character string. The name of the the targets in
#'   `targetcolname` (mapping information).
#' @param controllevelname A character string. The name of the the controls in
#'   `targetcolname` (mapping information).
#' @param targetcol A character string. The columname of `sample_metadata`
#'   holding the targets (default: "target"; caution: this should not
#'   be changed!).
#' @param targetname A character string. Name of the the targets,
#'   specified in the 'target' column of `sample_metadata` (default: "Target";
#'   caution: this should not be changed!).
#' @param controlname A character string. Name of the the controls,
#'   specified in the 'target' column of `sample_metadata` (default: "Control";
#'   caution: this should not be changed!).
#' @param use_rawdata A logical. Indicates, if raw the data is downloaded
#'   in CEL file format, uncompressed and subsequently normalized with a GCRMA
#'   normalization, or if the already normalized datasets from GEO are
#'   downloaded (default: FALSE).
#' @param setid A integer. Indicates, which set should be used by the software
#'   for subsequent analyses (default: 1).
#'
#' @export

geo_load_eset <- function(name,
                          datadir,
                          targetcolname,
                          targetlevelname,
                          controllevelname,
                          targetcol,
                          targetname,
                          controlname,
                          use_rawdata = FALSE,
                          setid = 1) {
  # original GEO data
  eset <- GEOquery::getGEO(
    name,
    destdir = datadir,
    AnnotGPL = TRUE
  )[[setid]]

  # rename targetcol
  colnames(Biobase::pData(eset))[which(
    colnames(Biobase::pData(eset)) == targetcolname
  )] <- targetcol

  if (!is.null(targetlevelname)) {
    # split levels, if more than one is provided for each targetlevelname
    # and controllevelname (recognized by '|||')
    if (grepl("(\\|){3}", targetlevelname)) { # nolint
      targetlevelname <- strsplit(
        targetlevelname, split = "|||", fixed = T
      )[[1]]
    }
    if (!any(targetlevelname %in% unique(
      as.character(Biobase::pData(eset)[[targetcol]])
    ))) {
      stop(paste0("\nLevel(s) provided with targetlevelname are ",
                  "not present in your data.\n"))
    }
  }

  if (!is.null(controllevelname)) {
    if (grepl("(\\|){3}", controllevelname)) { # nolint
      controllevelname <- strsplit(
        controllevelname, split = "|||", fixed = T
      )[[1]]
    }
    if (!any(controllevelname %in% unique(
      as.character(Biobase::pData(eset)[[targetcol]])
    ))) {
      stop(paste0("\nLevel(s) provided with controllevelname are ",
                  "not present in your data.\n"))
    }
  }

  # reduce pheno data to hold only cases of which we have
  # targetcolumname and controllevelname
  Biobase::pData(eset) <- Biobase::pData(eset)[which(
    Biobase::pData(eset)[[targetcol]] %in%
      c(targetlevelname, controllevelname)), ]

  # rename levels of targetcol
  if (!is.null(targetlevelname)) {
    t_levelnames <- rep(targetname, length(targetlevelname))
    names(t_levelnames) <- targetlevelname
  } else {
    t_levelnames <- NULL
  }

  if (!is.null(controllevelname)) {
    c_levelnames <- rep(controlname, length(controllevelname))
    names(c_levelnames) <- controllevelname
  } else {
    c_levelnames <- NULL
  }

  levelnames <- c(t_levelnames, c_levelnames)

  eset[[targetcol]] <- plyr::revalue(eset[[targetcol]], levelnames)

  if (isTRUE(use_rawdata)) {
    eset <- geo_use_raw_data(
      eset = eset,
      name = name,
      datadir = datadir
    )
  }

  # assign eset to the global env
  global_env_hack(
    key = name,
    val = eset,
    pos = 1L
  )

  # return eset as well
  return(eset)
}