56e1d09dbeac2c8233ef0ca36541e937a068c8aa ceisenhart Fri May 12 10:30:32 2017 -0700 Fixing up the usage statement, refs #19155 diff --git src/utils/expMatrixToBarchartBed/expMatrixToBarchartBed src/utils/expMatrixToBarchartBed/expMatrixToBarchartBed index 3e155c2..53e18f3 100755 --- src/utils/expMatrixToBarchartBed/expMatrixToBarchartBed +++ src/utils/expMatrixToBarchartBed/expMatrixToBarchartBed @@ -1,339 +1,340 @@ #!/usr/bin/env python2.7 # expMatrixToBarchartBed """ Generate a barChart bed6+5 file from a matrix, meta data, and coordinates. """ import os import sys import argparse import tempfile def parseArgs(args): """ Parse the command line arguments. """ parser= argparse.ArgumentParser(description = __doc__) parser.add_argument ("sampleFile", help = " Two column no header, the first column is the samples which should match " + \ "the matrix, the second is the grouping (cell type, tissue, etc)", type = argparse.FileType("r")) parser.add_argument ("matrixFile", help = " The input matrix file. The samples in the first row should exactly match the ones in " + \ "the sampleFile. The labels (ex ENST*****) in the first column should exactly match " + \ "the ones in the bed file.", type = argparse.FileType("r")) parser.add_argument ("bedFile", - help = " Bed4+2 format. File that maps the column labels from the matrix to coordinates. Tab " + \ - "separated; chr, start coord, end coord, label, strand, gene name. ", + help = " Bed6+1 format. File that maps the column labels from the matrix to coordinates. Tab " + \ + "separated; chr, start coord, end coord, label, score, strand, gene name. The score " + \ + "column is ignored.", action = "store") parser.add_argument ("outputFile", help = " The output file, bed 6+5 format. See the schema in kent/src/hg/lib/barChartBed.as. ", type =argparse.FileType("w")) # Optional arguments. parser.add_argument ("--groupOrderFile", help = " Optional file to define the group order, list the groups in a single column in " + \ "the order desired. The default ordering is alphabetical.", action = "store") parser.add_argument ("--useMean", help = " Calculate the group values using mean rather than median.", action = "store_true") parser.add_argument ("--verbose", help = " Show runtime messages.", action = "store_true") parser.set_defaults(verbose = False) parser.set_defaults(groupOrderFile = None) if (len(sys.argv) == 1): parser.print_help() exit(1) options = parser.parse_args() return options def median(lst): lst = sorted(lst) if len(lst) < 1: return None if len(lst) %2 == 1: return lst[((len(lst)+1)/2)-1] else: return float(sum(lst[(len(lst)/2)-1:(len(lst)/2)+1]))/2.0 def determineScore(tpmCutoffs, tpm): """ Cast the tpm to a score between 0-1000. Since there are only 9 visual blocks cast them to be in one of the 9 blocks. tpmCutoffs - A list of integers tpm - An integer """ count = 0 for val in tpmCutoffs: if (val > tpm): return count*111 count = count + 1 return 999 def floatRound(inFloat): """ Return a float that has at most 2 decimal places. """ beforeDecimal = True result = "" count = 0 for char in inFloat: if char is ".": beforeDecimal = False result += char continue if beforeDecimal: result += char else: if count >= 2: return result else: count += 1 result += char return result def condenseMatrixIntoBedCols(matrix, groupOrder, sampleToGroup, validTpms, bedLikeFile, useMean): """ Take an expression matrix and a dictionary that maps the samples to groups. Go through the expression matrix and calculate the average for each group, outputting it to an intermediate file as they are calculated. The intermediate file has three columns, the first is the average tpm for the entire gene, next is the number of groups and finally the average tpm for each group as a comma separated list. validTpms - An empty list of integers. sampleToGroup - A dictionary that maps string samples to string groups. matrix - An expression matrix, samples are the x rows, transcripts the y rows. bedLikeFile - An intermediate file, looks slightly like a bed. """ # Store some information on the bed file, most important is the order # of the 8th column. bedInfo = "" firstLine = True getBedInfo = True # Use the first line of the matrix and the sampleToGroup dict to create a dictionary that maps # the column to a group. columnToGroup = dict() # Go through the matrix line by line. The first line is used to build an index mapping columns # to group blocks, then for each line with TPM values merge the values based on group blocks. for line in matrix: splitLine = line.strip("\n").split() # The first line is the word 'transcript' followed by a list of the sample names. if firstLine: firstLine = False count = 1 firstCol = True for col in splitLine: if firstCol: firstCol = False continue group = sampleToGroup[col] columnToGroup.setdefault(count, group) count += 1 continue # Handle the tpm rows, calculating the average for each group per row. groupAverages = dict() groupCounts = dict() firstCol = True count = 1 for col in splitLine: if firstCol: firstCol = False continue if (useMean): # First time this group is seen, add it to the groupCounts dict. if (groupAverages.get(columnToGroup[count]) == None): groupAverages.setdefault(columnToGroup[count], float(col)) groupCounts.setdefault(columnToGroup[count], 1) # This group has already been seen, update the TPM average. else: groupCounts[columnToGroup[count]] += 1 # Average calculation normal = float(groupCounts[columnToGroup[count]]) newTpm = (float(col) * (1/normal)) oldTpm = (((normal - 1) / normal) * groupAverages[columnToGroup[count]]) groupAverages[columnToGroup[count]] = newTpm + oldTpm else: # First time this group is seen, add it to the groupCounts dict. if (groupAverages.get(columnToGroup[count]) == None): groupAverages.setdefault(columnToGroup[count], [float(col)]) groupCounts.setdefault(columnToGroup[count], 1) # This group has already been seen, update the TPM average. else: groupCounts[columnToGroup[count]] += 1 # Median preparation groupAverages[columnToGroup[count]].append(float(col)) count += 1 # Store some information on the bed file. Most important is the groupOrder. if getBedInfo: getBedInfo = False bedInfo += "#chr\tstart\tend\tname\tscore\tstrand\tname2\texpCount\texpScores;" if (groupOrder is not None): for group in open(groupOrder, "r"): bedInfo += group.strip("\n") + " " else: for key, value in sorted(groupAverages.iteritems()): bedInfo += key + " " bedInfo = bedInfo[:-1] + "\t_offset\t_lineLength" # Write out the transcript name, this is needed to join with coordinates later. bedLikeFile.write(splitLine[0] + "\t") # Create a list of the average scores per group. bedLine = "" # The fullAverage is used to assign a tpm score representative of the entire bed row. fullAverage = 0.0 count = 0.0 if (groupOrder is not None): for group in open(groupOrder, "r"): # Averages if (useMean): value = groupAverages[group.strip("\n")] else: value = median(groupAverages[group.strip("\n")]) bedLine = bedLine + "," + floatRound(str(value)) count += 1.0 fullAverage += value else: for key, value in sorted(groupAverages.iteritems()): if (useMean): bedLine = bedLine + "," + floatRound(str(value)) fullAverage += value else: bedLine = bedLine + "," + floatRound(str(median(value))) fullAverage += median(value) count += 1.0 # Create what will be columns 5, 7 and 8 of the final bed. bedLine = str(fullAverage/count) + "\t" + str(int(count)) + "\t" + bedLine[1:] + "\n" # If the fullAverage tpm is greater than 0 then consider it in the validTpm list. if (fullAverage > 0.0): validTpms.append((fullAverage/count)) # Write the bedLine to the intermediate bed-like file. bedLikeFile.write(bedLine) # Return the bedInfo so it can be printed right before the script ends. return bedInfo def expMatrixToBarchartBed(options): """ Convert the expression matrix into a barchart bed file. options - The command line options (file names, etc). Use the meta data to map the sample names to their groups, then create a dict that maps the columns to the groups. Go through the matrix line by line and get the median or average for each group. Print this to an intermediate file, then use the unix 'join' command to link with the coordinates file via the first matrix column. This creates a file with many of the bed fields just in the wrong order. Go through this file to re arrange the columns, check for and remove entries where chromsomes names include "_" and chr start > chr end. Finally run Max's bedJoinTabOffset to index the matrix adding the dataOffset and dataLen columns and creating a bed 6+5 file. """ # Create a dictionary that maps the sample names to their group. sampleToGroup = dict() count = 0 for item in options.sampleFile: count +=1 splitLine = item.strip("\n").split() if (len(splitLine) is not 2): print ("There was an error reading the sample file at line " + count) exit(1) sampleToGroup.setdefault(splitLine[0], splitLine[1]) # Use an intermediate file to hold the average values for each group. bedLikeFile = tempfile.NamedTemporaryFile( mode = "w+", bufsize = 1) # Keep a list of TPM scores greater than 0. This will be used later # to assign bed scores. validTpms = [] # Go through the matrix and condense it into a bed like file. Populate # the validTpms array and the bedInfo string. bedInfo = condenseMatrixIntoBedCols(options.matrixFile, options.groupOrderFile, sampleToGroup, \ validTpms, bedLikeFile, options.useMean) # Find the number which divides the list of non 0 TPM scores into ten blocks. tpmMedian = sorted(validTpms) blockSizes = len(tpmMedian)/10 # Create a list of the ten TPM values at the edge of each block. # These used to cast a TPM score to one of ten value between 0-1000. tpmCutoffs = [] for i in range(1,10): tpmCutoffs.append(tpmMedian[blockSizes*i]) # Sort the bed like file to prepare it for the join. sortedBedLikeFile = tempfile.NamedTemporaryFile( mode = "w+", bufsize = 1) cmd = "sort " + bedLikeFile.name + " > " + sortedBedLikeFile.name os.system(cmd) # Cut apart the coordinate bed to get the transcripts in the first column so it can be joined. coordBedPart1 = tempfile.NamedTemporaryFile( mode = "w+", bufsize = 1) cmd = "cut -f 4 " + options.bedFile + " > " + coordBedPart1.name os.system(cmd) coordBedPart2 = tempfile.NamedTemporaryFile( mode = "w+", bufsize = 1) cmd = "cut -f 1-3,5-7 " + options.bedFile + " > " + coordBedPart2.name os.system(cmd) # Sort the coordinate file to prepare it for the join. sortedCoords = tempfile.NamedTemporaryFile( mode = "w+", bufsize = 1) cmd = "paste " + coordBedPart1.name + " " + coordBedPart2.name + " | sort > " + sortedCoords.name os.system(cmd) # Join the bed-like file and the coordinate file. joinedFile = tempfile.NamedTemporaryFile(mode="w+", bufsize=1) cmd = "join " + sortedCoords.name + " " + sortedBedLikeFile.name + " | awk -v " + \ "OFS=\"\\t\" '$1=$1' > " + joinedFile.name os.system(cmd) # Go through the joined file and re arrange the columns creating a bed 6+2 file. # Also assign a scaled score 0 - 1000 to each tpm value. bedFile = tempfile.NamedTemporaryFile(mode="w+", bufsize=1) for line in joinedFile: splitLine = line.strip("\n").split() if ("_" in splitLine[0]): sys.stderr.write("This transcript " + splitLine[0] + " was dropped for having a '_' in the name.\n") continue # Ignore alt sequences. # Drop sequences where start is greater than end. if (float(splitLine[2]) > float(splitLine[3])): sys.stderr.write("This transcript " + splitLine[0] + " was dropped since chr end, " + \ splitLine[3] + ", is smaller than chr start, " + splitLine[2] + "\n.") continue chrom = splitLine[1] chrStart = splitLine[2] chrEnd = splitLine[3] name = splitLine[0] score = str(determineScore(tpmCutoffs, float(splitLine[7]))) strand = splitLine[5] name2 = splitLine[6] expCount = str(splitLine[7]) expScores = splitLine[8] bedLine = (chrom + "\t" + chrStart + "\t" + chrEnd + "\t" + name + "\t" \ + score + "\t" + strand + "\t" + name2 + "\t" + expCount + "\t" \ + expScores + "\n") bedFile.write(bedLine) # Run Max's indexing script indexedBedFile = tempfile.NamedTemporaryFile(mode="w+", bufsize=1) cmd = "bedJoinTabOffset " + options.matrixFile.name + " " + bedFile.name + " " + indexedBedFile.name os.system(cmd) # Prepend the bed info to the start of the file. cmd = "echo '" + bedInfo + "' > " + options.outputFile.name os.system(cmd) cmd = "cat " + indexedBedFile.name + " >> " + options.outputFile.name os.system(cmd) print ("The columns and order of the groups are; \n" + bedInfo) def main(args): """ Initialized options and calls other functions. """ options = parseArgs(args) expMatrixToBarchartBed(options) if __name__ == "__main__" : sys.exit(main(sys.argv))