Mercurial > repos > yufei-luo > s_mart
view SMART/Java/Python/plotCoverage.py @ 24:452e051f6562
Deleted selected files
author | m-zytnicki |
---|---|
date | Mon, 29 Apr 2013 03:28:24 -0400 |
parents | 94ab73e8a190 |
children |
line wrap: on
line source
#! /usr/bin/env python # # Copyright INRA-URGI 2009-2010 # # This software is governed by the CeCILL license under French law and # abiding by the rules of distribution of free software. You can use, # modify and/ or redistribute the software under the terms of the CeCILL # license as circulated by CEA, CNRS and INRIA at the following URL # "http://www.cecill.info". # # As a counterpart to the access to the source code and rights to copy, # modify and redistribute granted by the license, users are provided only # with a limited warranty and the software's author, the holder of the # economic rights, and the successive licensors have only limited # liability. # # In this respect, the user's attention is drawn to the risks associated # with loading, using, modifying and/or developing or reproducing the # software by the user in light of its specific status of free software, # that may mean that it is complicated to manipulate, and that also # therefore means that it is reserved for developers and experienced # professionals having in-depth computer knowledge. Users are therefore # encouraged to load and test the software's suitability as regards their # requirements in conditions enabling the security of their systems and/or # data to be ensured and, more generally, to use and operate it in the # same conditions as regards security. # # The fact that you are presently reading this means that you have had # knowledge of the CeCILL license and that you accept its terms. # import os, os.path, subprocess, glob, random from optparse import OptionParser from SMART.Java.Python.structure.Interval import Interval from SMART.Java.Python.structure.Transcript import Transcript from commons.core.parsing.ParserChooser import ParserChooser from SMART.Java.Python.misc.RPlotter import RPlotter from SMART.Java.Python.misc.Progress import Progress from commons.core.parsing.FastaParser import FastaParser strands = [-1, 1] colors = {-1: "blue", 1: "red", 0: "black"} colorLine = "black" def parseTargetField(field): strand = "+" splittedFieldSpace = field.split() splittedFieldPlus = field.split("+", 4) if len(splittedFieldSpace) == 3: id, start, end = splittedFieldSpace elif len(splittedFieldSpace) == 4: id, start, end, strand = splittedFieldSpace elif len(splittedFieldPlus) == 3: id, start, end = splittedFieldPlus elif len(splittedFieldPlus) == 4: id, start, end, strand = splittedFieldPlus else: raise Exception("Cannot parse Target field '%s'." % (field)) return (id, int(start), int(end), strand) class SimpleTranscript(object): def __init__(self, transcript1, transcript2, color = None): self.start = max(0, transcript1.getStart() - transcript2.getStart()) self.end = min(transcript2.getEnd() - transcript2.getStart(), transcript1.getEnd() - transcript2.getStart()) self.strand = transcript1.getDirection() * transcript2.getDirection() self.exons = [] for exon in transcript1.getExons(): if exon.getEnd() >= transcript2.getStart() and exon.getStart() <= transcript2.getEnd(): start = max(0, exon.getStart() - transcript2.getStart()) end = min(transcript2.getEnd() - transcript2.getStart(), exon.getEnd() - transcript2.getStart()) self.addExon(start, end, self.strand, color) def addExon(self, start, end, strand, color): exon = SimpleExon(start, end, strand, color) self.exons.append(exon) def getRScript(self, yOffset, height): rString = "" previousEnd = None for exon in sorted(self.exons, key=lambda exon: exon.start): if previousEnd != None: rString += "segments(%.1f, %.1f, %.1f, %.1f, col = \"%s\")\n" % (previousEnd, yOffset + height / 4.0, exon.start, yOffset + height / 4.0, colorLine) rString += exon.getRScript(yOffset, height) previousEnd = exon.end return rString class SimpleExon(object): def __init__(self, start, end, strand, color = None): self.start = start self.end = end self.strand = strand self.color = color def getRScript(self, yOffset, height): color = self.color if self.color != None else colors[self.strand] return "rect(%.1f, %.1f, %.1f, %.1f, col=\"%s\", border = \"%s\")\n" % (self.start, yOffset, self.end, yOffset + height / 2.0, color, colorLine) class Plotter(object): def __init__(self, seed, index, verbosity): self.seed = seed self.index = index self.verbosity = verbosity self.maxCoverage = 0 self.maxOverlap = 0 self.log = "" self.merge = False self.width = 1500 self.heigth = 1000 self.xLabel = "" self.yLabel = "" self.title = None self.absPath = os.getcwd() self.coverageDataFileName = "tmpFile_%d_%s.dat" % (seed, index) self.coverageScript = "" self.overlapScript = "" self.outputFileName = None def setOutputFileName(self, fileName): self.outputFileName = fileName def setTranscript(self, transcript): self.transcript = transcript self.name = transcript.getName() self.size = transcript.getEnd() - transcript.getStart() + 1 if self.title == None: self.title = self.name else: self.title += " " + self.name def setTitle(self, title): self.title = title + " " + self.name def setPlotSize(self, width, height): self.width = width self.height = height def setLabels(self, xLabel, yLabel): self.xLabel = xLabel self.yLabel = yLabel def setMerge(self, merge): self.merge = merge def setCoverageData(self, coverage): outputCoveragePerStrand = dict([strand, 0] for strand in strands) outputCoverage = 0 dataFile = open(os.path.abspath(self.coverageDataFileName), "w") for position in range(self.size+1): sumValue = 0 found = False dataFile.write("%d\t" % (position)) for strand in strands: value = coverage[strand].get(position, 0) sumValue += value dataFile.write("%d\t" % (value)) if value > 0: found = True outputCoveragePerStrand[strand] += 1 self.maxCoverage = max(self.maxCoverage, sumValue) dataFile.write("%d\n" % (sumValue)) if found: outputCoverage += 1 dataFile.close() self.log += "%s (%d nt):\n - both strands: %d (%.0f%%)\n - (+) strand: %d (%.0f%%)\n - (-) strand: %d (%.0f%%)\n" % (self.name, self.size, outputCoverage, float(outputCoverage) / self.size * 100, outputCoveragePerStrand[1], float(outputCoveragePerStrand[1]) / self.size * 100, outputCoveragePerStrand[-1], float(outputCoveragePerStrand[-1]) / self.size * 100) self.coverageScript += "data = scan(\"%s\", list(pos = -666, minus = -666, plus = -666, sumValue = -666), sep=\"\t\")\n" % (os.path.abspath(self.coverageDataFileName)) self.coverageScript += "lines(x = data$pos, y = data$minus, col = \"%s\")\n" % (colors[-1]) self.coverageScript += "lines(x = data$pos, y = data$plus, col = \"%s\")\n" % (colors[1]) self.coverageScript += "lines(x = data$pos, y = data$sumValue, col = \"%s\")\n" % (colors[0]) def setOverlapData(self, overlap): height = 1 self.maxOverlap = (len(overlap) + 1) * height thisElement = SimpleTranscript(self.transcript, self.transcript, "black") self.overlapScript += thisElement.getRScript(0, height) for cpt, transcript in enumerate(sorted(overlap, cmp=lambda c1, c2: c1.start - c2.start if c1.start != c2.start else c1.end - c2.end)): self.overlapScript += transcript.getRScript((cpt + 1) * height, height) def getFirstLine(self, suffix = None): return "png(file = \"%s_%s%s.png\", width = %d, height = %d, bg = \"white\")\n" % (self.outputFileName, self.name, "" if suffix == None or self.merge else "_%s" % (suffix), self.width, self.height) def getLastLine(self): return "dev.off()\n" def startR(self, fileName, script): scriptFile = open(fileName, "w") scriptFile.write(script) scriptFile.close() command = "R CMD BATCH %s" % (fileName) status = subprocess.call(command, shell=True) if status != 0: raise Exception("Problem with the execution of script file %s, status is: %s" % (fileName, status)) def plot(self): if self.merge: fileName = "%s_%d_%s.R" % (self.outputFileName, self.seed, self.index) plotLine = "plot(x = NA, y = NA, xlab=\"%s\", ylab=\"%s\", panel.first = grid(lwd = 1.0), xlim = c(0, %d), ylim = c(0, %d), cex.axis = 2, cex.lab = 2, cex.main=2, main = \"%s\")\n" % (self.xLabel, self.yLabel, self.size, max(self.maxCoverage, self.maxOverlap), self.title) script = self.getFirstLine() + plotLine + self.overlapScript + self.coverageScript + self.getLastLine() self.startR(fileName, script) else: fileName = "%s_%d_%s_overlap.R" % (self.outputFileName, self.seed, self.index) plotLine = "plot(x = NA, y = NA, xlab=\"%s\", ylab=\"%s\", panel.first = grid(lwd = 1.0), xlim = c(0, %d), ylim = c(0, %d), cex.axis = 2, cex.lab = 2, cex.main=2, main = \"%s\")\n" % (self.xLabel, self.yLabel, self.size, self.maxOverlap, self.title) script = self.getFirstLine("overlap") + plotLine + self.overlapScript + self.getLastLine() self.startR(fileName, script) fileName = "%s_%d_%s_coverage.R" % (self.outputFileName, self.seed, self.index) plotLine = "plot(x = NA, y = NA, xlab=\"%s\", ylab=\"%s\", panel.first = grid(lwd = 1.0), xlim = c(0, %d), ylim = c(0, %d), cex.axis = 2, cex.lab = 2, cex.main=2, main = \"%s\")\n" % (self.xLabel, self.yLabel, self.size, self.maxCoverage, self.title) script = self.getFirstLine("coverage") + plotLine + self.coverageScript + self.getLastLine() self.startR(fileName, script) class PlotParser(object): def __init__(self, verbosity): self.verbosity = verbosity self.parsers = [None, None] self.sequenceParser = None self.seed = random.randint(0, 10000) self.title = "" self.merge = False def __del__(self): for fileName in glob.glob("tmpFile_%d*.dat" % (self.seed)): os.remove(fileName) for fileName in glob.glob("%s*.R" % (os.path.abspath(self.outputFileName))): os.remove(fileName) for fileName in glob.glob("%s*.Rout" % (os.path.abspath(self.outputFileName))): os.remove(fileName) def addInput(self, inputNb, fileName, fileFormat): if fileName == None: return chooser = ParserChooser(self.verbosity) chooser.findFormat(fileFormat) self.parsers[inputNb] = chooser.getParser(fileName) if inputNb == 0: self.parsers[1] = self.parsers[0] def addSequence(self, fileName): if fileName == None: return self.sequenceParser = FastaParser(fileName, self.verbosity) def setOutput(self, fileName): self.outputFileName = fileName def setPlotSize(self, width, height): self.width = width self.height = height def setLabels(self, xLabel, yLabel): self.xLabel = xLabel self.yLabel = yLabel def setTitle(self, title): self.title = title def setMerge(self, merge): self.merge = merge def initializeDataFromSequences(self): self.sizes = {} self.coverage = {} self.overlap = {} for region in self.sequenceParser.getRegions(): self.sizes[region] = self.sequenceParser.getSizeOfRegion(region) self.coverage[region] = {} self.overlap[region] = [] for strand in strands: self.coverage[region][strand] = {} self.coverage[region][strand][1] = 0 self.coverage[region][strand][self.sizes[region]] = 0 def initializeDataFromTranscripts(self): self.coverage = dict([i, None] for i in range(self.parsers[1].getNbTranscripts())) self.overlap = dict([i, None] for i in range(self.parsers[1].getNbTranscripts())) self.sizes = dict([i, 0] for i in range(self.parsers[1].getNbTranscripts())) progress = Progress(self.parsers[1].getNbTranscripts(), "Reading regions", self.verbosity) for cpt, transcript in enumerate(self.parsers[1].getIterator()): self.coverage[cpt] = {} self.overlap[cpt] = [] for strand in strands: self.coverage[cpt][strand] = {} self.coverage[cpt][strand][0] = 0 self.coverage[cpt][strand][transcript.getEnd() - transcript.getStart()] = 0 for exon in transcript.getExons(): self.sizes[cpt] += exon.getSize() progress.inc() progress.done() def initialize(self): if self.sequenceParser == None: self.initializeDataFromTranscripts() else: self.initializeDataFromSequences() def computeCoverage(self, transcript1, transcript2, id): strand = transcript1.getDirection() * transcript2.getDirection() for exon1 in transcript1.getExons(): for exon2 in transcript2.getExons(): if exon1.overlapWith(exon2): for position in range(max(exon1.getStart(), exon2.getStart()), min(exon1.getEnd(), exon2.getEnd()) + 1): relativePosition = position - transcript2.getStart() + 1 self.coverage[id][strand][relativePosition] = self.coverage[id][strand].get(relativePosition, 0) + 1 def computeOverlap(self, transcript1, transcript2, id): simpleTranscript = SimpleTranscript(transcript1, transcript2) self.overlap[id].append(simpleTranscript) def compute2TranscriptFiles(self): progress = Progress(self.parsers[1].getNbTranscripts(), "Comparing regions", self.verbosity) for cpt2, transcript2 in enumerate(self.parsers[1].getIterator()): for transcript1 in self.parsers[0].getIterator(): if transcript1.overlapWithExon(transcript2): self.computeCoverage(transcript1, transcript2, cpt2) self.computeOverlap(transcript1, transcript2, cpt2) progress.inc() progress.done() def extractReferenceQueryMapping(self, mapping): queryTranscript = mapping.getTranscript() referenceTranscript = Transcript() referenceTranscript.setChromosome(queryTranscript.getChromosome()) referenceTranscript.setName(queryTranscript.getChromosome()) referenceTranscript.setDirection("+") referenceTranscript.setEnd(self.sizes[queryTranscript.getChromosome()]) referenceTranscript.setStart(1) return (referenceTranscript, queryTranscript) def extractReferenceQuery(self, inputTranscript): if "Target" not in inputTranscript.getTagNames(): raise Exception("Cannot extract Target field in line '%s'." % (inputTranscript)) id, start, end, strand = parseTargetField(inputTranscript.getTagValue("Target")) if id not in self.sizes: raise Exception("Target id '%s' of transcript '%s' does not correspond to anything in FASTA file." % (id, inputTranscript)) referenceTranscript = Transcript() referenceTranscript.setChromosome(id) referenceTranscript.setName(id) referenceTranscript.setDirection("+") referenceTranscript.setEnd(self.sizes[id]) referenceTranscript.setStart(1) queryTranscript = Transcript() queryTranscript.setChromosome(id) queryTranscript.setName(id) queryTranscript.setStart(start) queryTranscript.setEnd(end) queryTranscript.setDirection(strand) if inputTranscript.getNbExons() > 1: factor = float(end - start) / (inputTranscript.getEnd() - inputTranscript.getStart()) for exon in inputTranscript.getExons(): newExon = Interval() newExon.setChromosome(id) newExon.setDirection(strand) if "Target" in inputTranscript.getTagNames(): id, start, end, strand = parseTargetField(exon.getTagValue("Target")) newExon.setStart(start) newExon.setEnd(end) else: newExon.setStart(int(round((exon.getStart() - inputTranscript.getStart()) * factor)) + start) newExon.setEnd( int(round((exon.getEnd() - inputTranscript.getStart()) * factor)) + start) queryTranscript.addExon(newExon) return (referenceTranscript, queryTranscript) def compute1TranscriptFiles(self): progress = Progress(self.parsers[1].getNbItems(), "Comparing regions", self.verbosity) for transcript in self.parsers[1].getIterator(): if transcript.__class__.__name__ == "Mapping": referenceTranscript, queryTranscript = self.extractReferenceQueryMapping(transcript) else: referenceTranscript, queryTranscript = self.extractReferenceQuery(transcript) self.computeCoverage(queryTranscript, referenceTranscript, referenceTranscript.getName()) self.computeOverlap(queryTranscript, referenceTranscript, referenceTranscript.getName()) progress.inc() progress.done() def compute(self): if self.sequenceParser == None: self.compute2TranscriptFiles() else: self.compute1TranscriptFiles() def plotTranscript(self, index, transcript): plotter = Plotter(self.seed, index, self.verbosity) plotter.setOutputFileName(self.outputFileName) plotter.setTranscript(transcript) plotter.setTitle(self.title) plotter.setLabels(self.xLabel, self.yLabel) plotter.setPlotSize(self.width, self.height) plotter.setCoverageData(self.coverage[index]) plotter.setOverlapData(self.overlap[index]) plotter.setMerge(self.merge) plotter.plot() output = plotter.log return output def plot1TranscriptFile(self): self.outputCoverage = {} self.outputCoveragePerStrand = {} output = "" progress = Progress(len(self.sequenceParser.getRegions()), "Plotting regions", self.verbosity) for cpt2, region in enumerate(self.sequenceParser.getRegions()): transcript = Transcript() transcript.setName(region) transcript.setDirection("+") transcript.setEnd(self.sizes[region]) transcript.setStart(1) output += self.plotTranscript(region, transcript) progress.inc() progress.done() if self.verbosity > 0: print output def plot2TranscriptFiles(self): self.outputCoverage = [0] * self.parsers[1].getNbTranscripts() self.outputCoveragePerStrand = [None] * self.parsers[1].getNbTranscripts() for cpt in range(self.parsers[1].getNbTranscripts()): self.outputCoveragePerStrand[cpt] = dict([strand, 0] for strand in strands) progress = Progress(self.parsers[1].getNbTranscripts(), "Plotting regions", self.verbosity) output = "" for cpt2, transcript2 in enumerate(self.parsers[1].getIterator()): output += self.plotTranscript(cpt2, transcript2) progress.inc() progress.done() if self.verbosity > 0: print output def plot(self): if self.sequenceParser == None: self.plot2TranscriptFiles() else: self.plot1TranscriptFile() def start(self): self.initialize() self.compute() self.plot() if __name__ == "__main__": # parse command line description = "Plot Coverage v1.0.1: Plot the coverage of the first data with respect to the second one. [Category: Visualization]" parser = OptionParser(description = description) parser.add_option("-i", "--input1", dest="inputFileName1", action="store", type="string", help="input file 1 [compulsory] [format: file in transcript or mapping format given by -f]") parser.add_option("-f", "--inputFormat1", dest="inputFormat1", action="store", type="string", help="format of input file 1 [compulsory] [format: transcript or mapping file format]") parser.add_option("-j", "--input2", dest="inputFileName2", action="store", type="string", help="input file 2 [compulsory] [format: file in transcript format given by -g]") parser.add_option("-g", "--inputFormat2", dest="inputFormat2", action="store", type="string", help="format of input file 2 [compulsory] [format: transcript file format]") parser.add_option("-q", "--sequence", dest="inputSequence", action="store", default=None, type="string", help="input sequence file [format: file in FASTA format] [default: None]") parser.add_option("-o", "--output", dest="outputFileName", action="store", type="string", help="output file [compulsory] [format: output file in PNG format]") parser.add_option("-w", "--width", dest="width", action="store", default=1500, type="int", help="width of the plots (in px) [format: int] [default: 1500]") parser.add_option("-e", "--height", dest="height", action="store", default=1000, type="int", help="height of the plots (in px) [format: int] [default: 1000]") parser.add_option("-t", "--title", dest="title", action="store", default="", type="string", help="title of the plots [format: string]") parser.add_option("-x", "--xlab", dest="xLabel", action="store", default="", type="string", help="label on the x-axis [format: string]") parser.add_option("-y", "--ylab", dest="yLabel", action="store", default="", type="string", help="label on the y-axis [format: string]") parser.add_option("-p", "--plusColor", dest="plusColor", action="store", default="red", type="string", help="color for the elements on the plus strand [format: string] [default: red]") parser.add_option("-m", "--minusColor", dest="minusColor", action="store", default="blue", type="string", help="color for the elements on the minus strand [format: string] [default: blue]") parser.add_option("-s", "--sumColor", dest="sumColor", action="store", default="black", type="string", help="color for 2 strands coverage line [format: string] [default: black]") parser.add_option("-l", "--lineColor", dest="lineColor", action="store", default="black", type="string", help="color for the lines [format: string] [default: black]") parser.add_option("-1", "--merge", dest="merge", action="store_true", default=False, help="merge the 2 plots in 1 [format: boolean] [default: false]") parser.add_option("-D", "--directory", dest="working_Dir", action="store", default=os.getcwd(), type="string", help="the directory to store the results [format: directory]") parser.add_option("-v", "--verbosity", dest="verbosity", action="store", default=1, type="int", help="trace level [format: int]") (options, args) = parser.parse_args() colors[1] = options.plusColor colors[-1] = options.minusColor colors[0] = options.sumColor colorLine = options.lineColor pp = PlotParser(options.verbosity) pp.addInput(0, options.inputFileName1, options.inputFormat1) pp.addInput(1, options.inputFileName2, options.inputFormat2) pp.addSequence(options.inputSequence) pp.setOutput(options.outputFileName if os.path.isabs(options.outputFileName) else os.path.join(options.working_Dir, options.outputFileName)) pp.setPlotSize(options.width, options.height) pp.setLabels(options.xLabel, options.yLabel) pp.setTitle(options.title) pp.setMerge(options.merge) pp.start()