R/calculateFC_public.R

In anschue/toxprofileR: Functions for Retrieving Dynamic Toxicogenomic Fingerprints

#' Calculate logFC of public microarray data
#'
#' @param data EList
#' @param comparison data.frame describing the comparison for logFC
#'
#' @return dataframe of logFCs
#' @export
calc_logfc_public <- function(data, comparison) {
  treatment <- as.character(data$targets$Comp_SubstanceName_Trivial) == as.character(comparison$Comp_SubstanceName_Trivial.treatment) &
    (data$targets$Exp_Conc == comparison$Exp_Conc.treatment | is.na(comparison$Exp_Conc.treatment)) &
    (data$targets$Exp_Messzeit_hpf == comparison$Exp_Messzeit_hpf | is.na(comparison$Exp_Messzeit_hpf)) &
    (data$targets$Exp_Expositionsstart_hpf == comparison$Exp_Expositionsstart_hpf | is.na(comparison$Exp_Expositionsstart_hpf)) &
    (data$targets$Exp_Expositionsstop_hpf == comparison$Exp_Expositionsstop_hpf | is.na(data$targets$Exp_Expositionsstop_hpf))


  control <- as.character(data$targets$Comp_SubstanceName_Trivial) == as.character(comparison$Comp_SubstanceName_Trivial.control) &
    (data$targets$Exp_Conc == comparison$Exp_Conc.control | is.na(comparison$Exp_Conc.control)) &
    (data$targets$Exp_Messzeit_hpf == comparison$Exp_Messzeit_hpf | is.na(comparison$Exp_Messzeit_hpf))

  if (sum(treatment) > 1 & sum(control) > 1) {

    ## take only probes for each gene with highest IQR
    IQRs <- apply(data$E, 1, IQR, na.rm = T)
    data <- data[order(IQRs, decreasing = T), ]
    data <- data[!duplicated(data$genes$ensembl_gene_id), ]
    data <- data[!is.na(data$genes$ensembl_gene_id), ]

    mean_logFC <- apply(data$E, MARGIN = 1, function(probe) {
      mean(probe[treatment]) - mean(probe[control])
    })
    logFCframe <- data.frame(row.names = data$genes$ensembl_gene_id, logFC = mean_logFC)

    colnames(logFCframe) <- comparison$SampleName.treatment
    return(logFCframe)
  } else {
    logFCframe <- NA
    message("no replication - no FC")
    return(logFCframe)
  }
}