view CHM.R @ 38:605ec840a06b draft

Uploaded

### This method generates a row and column ordering given an input matrix and ordering methods.
###
### matrixData - numeric matrix 
### rowOrderMethod - Hierarchical, Original, Random
### rowDistanceMeasure - For clustering, distance measure. May be: euclidean, binary, manhattan, maximum, canberra, minkowski, or correlation.
### rowAgglomerationMethod - For clustering, agglomeration method.  May be:  'average' for Average Linkage, 'complete' for Complete Linkage,
###                                                                          'single' for Single Linkage, 'ward', 'mcquitty', 'median', or 'centroid'.
### colOrderMethod 
### colDistanceMeasure
### colAgglomerationMethod
### rowOrderFile - output file of order of rows 
### rowDendroFile - output file of row dendrogram  
### colOrderFile - output file of order of cols
### colDendroFile - output file of col dendrogram
### rowCut - For rows the number of classifications to automatically generate based on dendrogram into a classification file.  0 for turned off.
### colCut - For columns the number of classifications to automatically generate based on dendrogram into a classification file.  0 for turned off.

performDataOrdering<-function(dataFile, rowOrderMethod, rowDistanceMeasure, rowAgglomerationMethod, colOrderMethod, colDistanceMeasure, colAgglomerationMethod,rowOrderFile, colOrderFile, rowDendroFile, colDendroFile, rowCut, colCut)
{ 
   dataMatrix = read.table(dataFile, header=TRUE, sep = "\t", check.names = FALSE, row.names = 1, as.is=TRUE, na.strings=c("NA","N/A","-","?"))
   rowOrder <-  createOrdering(dataMatrix, rowOrderMethod, "row", rowDistanceMeasure, rowAgglomerationMethod)  
   if (rowOrderMethod == "Hierarchical") {
      writeHCDataTSVs(rowOrder, rowDendroFile, rowOrderFile)
   }

   colOrder <-  createOrdering(dataMatrix, colOrderMethod, "col", colDistanceMeasure, colAgglomerationMethod)  
   if (colOrderMethod == "Hierarchical") {
      writeHCDataTSVs(colOrder, colDendroFile, colOrderFile)
      writeHCCut(colOrder, colCut, paste(colOrderFile,".cut", sep=""))
   }
}

#creates output files for hclust ordering
writeHCDataTSVs<-function(uDend, outputHCDataFileName, outputHCOrderFileName)
{
   data<-cbind(uDend$merge, uDend$height, deparse.level=0)
   colnames(data)<-c("A", "B", "Height")
   write.table(data, file = outputHCDataFileName, append = FALSE, quote = FALSE, sep = "\t", row.names=FALSE)
 
   data=matrix(,length(uDend$labels),2);
   for (i in 1:length(uDend$labels)) {
   	  print(uDend$labels[i])
      data[i,1] = uDend$labels[i];
      data[i,2] = which(uDend$order==i);
   }
   colnames(data)<-c("Id", "Order")
   write.table(data, file = outputHCOrderFileName, append = FALSE, quote = FALSE, sep = "\t", row.names=FALSE)
}

#creates a classification file based on user specified cut of dendrogram
writeHCCut<-function(uDend, cutNum, outputCutFileName)
{
   if (cutNum < 2) {
      return()
   }
   print (paste("Writing cut file ", outputCutFileName))
   cut <- cutree(uDend, cutNum);
   id <- names(cut);
   data=matrix(,length(cut),2);
   for (i in 1:length(cut)) {
      data[i,1] = id[i];
      data[i,2] = sprintf("Cluster %d", cut[i]);
   }

   write.table(data, file = outputCutFileName, append = FALSE, quote = FALSE, sep = "\t", row.names=FALSE, col.names = FALSE);
}


createOrdering<-function(matrixData, orderMethod, direction, distanceMeasure, agglomerationMethod)
{
  ordering <- NULL

  if (orderMethod == "Hierarchical")
  {

    # Compute dendrogram for "Distance Metric"
    distVals <- NULL
    if(direction=="row") {
      if (distanceMeasure == "correlation") {
        geneGeneCor <- cor(t(matrixData), use="pairwise")
        distVals <- as.dist((1-geneGeneCor)/2)
      } else {
        distVals <- dist(matrixData, method=distanceMeasure)
      }
    } else { #column
      if (distanceMeasure == "correlation") {
        geneGeneCor <- cor(matrixData, use="pairwise")
        distVals <- as.dist((1-geneGeneCor)/2)
      } else {
        distVals <- dist(t(matrixData), method=distanceMeasure)
      }
    }

#    if (agglomerationMethod == "ward") {
#      ordering <- hclust(distVals * distVals, method="ward.D2")
#    } else {
      ordering <- hclust(distVals, method=agglomerationMethod)
#    }
  }
  else if (orderMethod == "Random")
  {
    if(direction=="row") {
       headerList <- rownames(matrixData)
       ordering <- sample(headerList, length(headerList)) 
    } else {
       headerList <- colnames(matrixData)
       ordering <- sample(headerList, length(headerList)) 
    }
  }
  else if (orderMethod == "Original")
  {
    if(direction=="row") {
       ordering <- rownames(matrixData) 
    } else {
       ordering <- colnames(matrixData) 
    }
  } else {
    stop("createOrdering -- failed to find ordering method")
  }
  return(ordering)
}
### Initialize command line arguments and call performDataOrdering

options(warn=-1)

args = commandArgs(TRUE)

performDataOrdering(dataFile=args[1], rowOrderMethod=args[2], rowDistanceMeasure=args[3], rowAgglomerationMethod=args[4], colOrderMethod=args[5], colDistanceMeasure=args[6], colAgglomerationMethod=args[7],rowOrderFile=args[8], colOrderFile=args[9], rowDendroFile=args[10], colDendroFile=args[11], rowCut=args[12], colCut=args[13])

#suppressWarnings(performDataOrdering(dataFile=args[1], rowOrderMethod=args[2], rowDistanceMeasure=args[3], rowAgglomerationMethod=args[4], colOrderMethod=args[5], colDistanceMeasure=args[6], colAgglomerationMethod=args[7],rowOrderFile=args[8], colOrderFile=args[9], rowDendroFile=args[10], colDendroFile=args[11]))