R/Recluster.R

In EDePasquale/DoubletDecon: A Package to Identify Doublets and Doublet Clusters Based on Deconvolution Analysis and Unique Gene Expression

#' Recluster
#'
#' This function reclusters doublets and non doublets seperately based on deconvolution analysis and returns a new expression file, groups file, and DeconFreq table for downstream analyses.
#' @param isADoublet isADoublet data.frame from Is_A_Doublet.
#' @param data Processed data from Clean_Up_Input (or Remove_Cell_Cycle).
#' @param groups Processed groups file from Clean_Up_Input.
#' @param log_file_name used for saving run notes to log file
#' @return newData2 - processed expression and groups file, reordered, with correct new cluster numbers.
#' @return decon - DeconCalledFreq table with all doublets 100 percent doublet and all non doublets at 0 percent doublet frequency.
#' @keywords recluster
#' @export

Recluster<-function(isADoublet, data, groups, log_file_name){

  #Get list of doublet samples
  doubletCells=row.names(subset(isADoublet, isADoublet==T))
  notDoubletCells=row.names(subset(isADoublet, isADoublet==F))

  #Make new expression tables for doublets and non-doublets individually
  doubletCellsData=data[2:nrow(data),colnames(data) %in% doubletCells]
  notDoubletCellsData=data[2:nrow(data),colnames(data) %in% notDoubletCells]
  doubletCellsData=t(as.matrix(doubletCellsData))
  notDoubletCellsData=t(as.matrix(notDoubletCellsData))

  #Use "clustering" by deconvolution predicted groups.
  nondoublets=groups[notDoubletCells,] #non doublets groups
  doublets=NULL #doublets groups
  isADoublet2=subset(isADoublet, isADoublet==TRUE)
  isADoublet2=isADoublet2$Cell_Types
  partialGroups=cbind(isADoublet2,isADoublet2)
  row.names(partialGroups)=row.names(subset(isADoublet, isADoublet==TRUE))
  colnames(partialGroups)=colnames(groups)
  doublets=partialGroups[order(partialGroups[,1]),]
  isADoublet3=doublets[,2]
  doublets[,1]=as.integer(as.factor(doublets[,1]))
  doublets[,2]=as.integer(as.factor(doublets[,2]))
  doublets[,1]=as.integer(doublets[,1])+length(unique(nondoublets[,1]))
  doublets[,2]=isADoublet3

  #merge these files together and create new groups classification
  newGroups=rbind(nondoublets, doublets)

  #Make Decon frequency table to return with 0% and 100%
  uniqueClusters=as.character(unique(newGroups[,2]))
  DeconCalledFreq=as.data.frame(matrix(nrow=length(uniqueClusters), ncol=1), row.names = uniqueClusters)
  DeconCalledFreq[1:length(unique(nondoublets[,1])),1]=0
  DeconCalledFreq[(length(unique(nondoublets[,1]))+1):nrow(DeconCalledFreq),1]=100

  #create new reordered expression file for return
  newData=data[,match(row.names(newGroups), colnames(data)) ]
  if(colnames(data)[1] %in% "row_clusters.flat" || colnames(data)[1] %in% "row_clusters-flat"){
    newData2=Clean_Up_Input(newData, newGroups, data[2:nrow(data), 1], log_file_name = log_file_name)
  }else{
    newData2=Clean_Up_Input(newData, newGroups, rowClusters = NULL , log_file_name = log_file_name)
  }

  return(list(newData2=newData2, decon=DeconCalledFreq))

}