R/SampleSelector.R

In JRC-COMBINE/FORESEE: A Tool for the Systematic Comparison of Translational Drug Response Modeling Pipelines

#' SampleSelector
#'
#' The Sample Selector returns features of only those samples that have a drug response for the drug specified by the user's input for DrugName and are available for all chosen input data types. If a specific tissue is selected, the samples are reduced to cell lines of that tissue only.
#'
#' @param TrainObject Object that contains all data needed to train a model, including molecular data (such as gene expression, mutation, copy number variation, methylation, cancer type, etc. ) and drug response data
#' @param TrainingTissue Tissue type that the cell lines of the TrainObject should be of, such as "lung". Default should be "all" for pancancer analysis.
#' @param InputDataTypes Data types of the TrainObject that are to be used to train the model, such as "GeneExpression", "Mutation", "CopyNumberVariation", "Methylation", "CancerType", etc.
#' @return \item{TrainObject}{The TrainObject with only those samples that have the specified drug response and are available for all chosen input data types.}
#' @examples
#' SampleSelector(GDSC, "pancreas", "GeneExpression")
#' SampleSelector(GDSC, "all", c("GeneExpression","Mutation"))
#' @export

#########################
# This file is part of the FORESEE R-package
# File authors: Lisa-Katrin Turnhoff <turnhoff@combine.rwth-aachen.de> and Ali Hadizadeh Esfahani <hadizadeh@combine.rwth-aachen.de>
# Distributed under the GNU General Public License v3.0.(http://www.gnu.org/licenses/gpl-3.0.html)
#########################

SampleSelector<- function(TrainObject,TrainingTissue, InputDataTypes){

  #################################################################################################################################
  # Check user's choice: Use all cell lines for training or only those of a specific tissue
  #################################################################################################################################
  # Check if the user wants to train on all cell lines (pan-cancer analysis)
  if (TrainingTissue=="all"){

    if (("GeneExpression" %in% InputDataTypes) == TRUE){
      CommonSamples <- colnames(TrainObject$GeneExpression)[colnames(TrainObject$GeneExpression) %in% names(TrainObject$DrugResponse)]
      TrainObject$GeneExpression <- TrainObject$GeneExpression[,CommonSamples]
      TrainObject$DrugResponse <- TrainObject$DrugResponse[CommonSamples]
    }
    if (("Mutation" %in% InputDataTypes) == TRUE){
      CommonSamples <- colnames(TrainObject$Mutation)[colnames(TrainObject$Mutation) %in% names(TrainObject$DrugResponse)]
      TrainObject$Mutation <- TrainObject$Mutation[,CommonSamples]
      TrainObject$DrugResponse <- TrainObject$DrugResponse[CommonSamples]
    }
    if (("Methylation" %in% InputDataTypes) == TRUE){
      CommonSamples <- colnames(TrainObject$Methylation)[colnames(TrainObject$Methylation) %in% names(TrainObject$DrugResponse)]
      TrainObject$Methylation <- TrainObject$Methylation[,CommonSamples]
      TrainObject$DrugResponse <- TrainObject$DrugResponse[CommonSamples]
    }
    if (("CopyNumberVaration" %in% InputDataTypes) == TRUE){
      CommonSamples <- colnames(TrainObject$CNVGain)[colnames(TrainObject$CNVGain) %in% names(TrainObject$DrugResponse)]
      TrainObject$CNVGain <- TrainObject$CNVGain[,CommonSamples]
      CommonSamples <- colnames(TrainObject$CNVLoss)[colnames(TrainObject$CNVLoss) %in% names(TrainObject$DrugResponse)]
      TrainObject$CNVLoss <- TrainObject$CNVLoss[,CommonSamples]
      TrainObject$DrugResponse <- TrainObject$DrugResponse[CommonSamples]
    }
    if (("CancerType" %in% InputDataTypes) == TRUE){

    CommonSamples <- colnames(TrainObject$CancerType)[colnames(TrainObject$CancerType) %in% names(TrainObject$DrugResponse)]
    TrainObject$CancerType <- TrainObject$CancerType[,CommonSamples]
    TrainObject$DrugResponse <- TrainObject$DrugResponse[CommonSamples]
    }
  }

  # Check if the user wants to train on cell linesof a specific tissue
  else if ((TrainingTissue %in% TrainObject$TissueInfo$Site)==TRUE){

    CellsofTissue <-TrainObject$TissueInfo$Line[TrainObject$TissueInfo$Site==TrainingTissue]
    RelevantSamples <- CellsofTissue[CellsofTissue %in% names(TrainObject$DrugResponse)]

    if (("GeneExpression" %in% InputDataTypes) == TRUE){
      TissueSamples <- colnames(TrainObject$GeneExpression)[colnames(TrainObject$GeneExpression) %in% RelevantSamples]
      TrainObject$GeneExpression <- TrainObject$GeneExpression[,TissueSamples]
      TrainObject$DrugResponse <- TrainObject$DrugResponse[TissueSamples]
    }
    if (("Mutation" %in% InputDataTypes) == TRUE){
      TissueSamples <- colnames(TrainObject$Mutation)[colnames(TrainObject$Mutation) %in% RelevantSamples]
      TrainObject$Mutation <- TrainObject$Mutation[,TissueSamples]
      TrainObject$DrugResponse <- TrainObject$DrugResponse[TissueSamples]
    }
    if (("Methylation" %in% InputDataTypes) == TRUE){
      TissueSamples <- colnames(TrainObject$Methylation)[colnames(TrainObject$Methylation) %in% RelevantSamples]
      TrainObject$Methylation <- TrainObject$Methylation[,TissueSamples]
      TrainObject$DrugResponse <- TrainObject$DrugResponse[TissueSamples]
    }
    if (("CopyNumberVaration" %in% InputDataTypes) == TRUE){
      TissueSamples <- colnames(TrainObject$CNVGain)[colnames(TrainObject$CNVGain) %in% RelevantSamples]
      TrainObject$CNVGain <- TrainObject$CNVGain[,TissueSamples]
      TissueSamples <- colnames(TrainObject$CNVLoss)[colnames(TrainObject$CNVLoss) %in% RelevantSamples]
      TrainObject$CNVLoss <- TrainObject$CNVLoss[,TissueSamples]
      TrainObject$DrugResponse <- TrainObject$DrugResponse[TissueSamples]
    }
    if (("CancerType" %in% InputDataTypes) == TRUE){

      TissueSamples <- colnames(TrainObject$CancerType)[colnames(TrainObject$CancerType) %in% RelevantSamples]
      TrainObject$CancerType <- TrainObject$CancerType[,TissueSamples]
      TrainObject$DrugResponse <- TrainObject$DrugResponse[TissueSamples]
    }
  }

  # Stop if specified tissue is not part of the tissue types listed in the Foresee TrainObject
  else{
    stop(paste0("The specified tissue '",TrainingTissue,"' is not included as a tissue type in the ForeseeObject."))
  }


  #################################################################################################################################
  # Identify those samples that are characterized for all chosen input data types
  #################################################################################################################################

  if (length(InputDataTypes)>1){

    JointSamples<-c()

    for (i in 1:length(InputDataTypes)){

      FeatureType <- InputDataTypes[i]
      JointSamples[[paste0("Samples_Features_",i)]] <- as.list(colnames(TrainObject[[FeatureType]]))
    }

    JointSamples <- as.character(Reduce(intersect, JointSamples))

    for (i in 1:length(InputDataTypes)){
      FeatureType <- InputDataTypes[i]
      TrainObject[[FeatureType]] <- TrainObject[[FeatureType]][,JointSamples]
    }

    TrainObject$DrugResponse <- TrainObject$DrugResponse[JointSamples]

    if (!length(JointSamples)>1){ # Stop if samples for the selected input data types do not overlap
      stop(paste0("The subsets of samples for the specified input data types '",InputDataTypes,"' do not overlap."))
    }
  }
  # Returning the new object
  return(TrainObject)
}