R/impute_lfq.R

In MassDynamics/lfq_processing: Processes Proteomics Files for Quantification.

#' impute LFQ data
#'
#' This function does MNAR imputation
#'
#' @param myQuantDT quantification data
#' @param id_type the id column of myQuantDT that indicates
#' @param int_type the column of myQuantDT to be imputed
#' @param f_imputStDev The Standard Deviation parameter for MNAR Imputation
#' @param f_imputePosition The Position parameter for MNAR Imputation
#' @return quantification data with missing values imputed
#' @export impute_lfq
impute_lfq <- function(myQuantDT,
                       id_type,
                       int_type,
                       f_imputeStDev = imputeStDev,
                       f_imputePosition = imputePosition) {

  myQuantDT[, int_impute := get(int_type)]
  myQuantDT <- myQuantDT[, colnames(myQuantDT) != int_type, with = F]
  impute.StdDev <-
    f_imputeStDev * myQuantDT[Imputed == 0, sd(int_impute)]
  impute.position <-
    myQuantDT[Imputed == 0, mean(int_impute)] - f_imputePosition * myQuantDT[Imputed == 0, sd(int_impute)]
  set.seed(255)
  myQuantDT[Imputed == 1, int_impute := rnorm(.N, (impute.position), (impute.StdDev))]

  # Add columns for centered and centered + scaled intensities
  myQuantDT[, nRLE := scale(int_impute, center = TRUE, scale = FALSE), by = .(get(id_type))]
  myQuantDT[, z_norm := scale(int_impute, center = TRUE, scale = TRUE), by = .(get(id_type))]

  setnames(myQuantDT, "int_impute", int_type)

  return(myQuantDT)
}