Mercurial > repos > yufei-luo > s_mart
view SMART/Java/Python/getLetterDistribution.py @ 64:783e6ed4eb66
Minor bug correction.
Added casts to str in Galaxy XML files. Also closed the writer in the Python script "changeTagName."
author | m-zytnicki |
---|---|
date | Mon, 19 Oct 2015 14:16:44 +0200 |
parents | 769e306b7933 |
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. # """Get the size distribution of a Fasta / BED file""" import os from optparse import OptionParser from commons.core.parsing.FastaParser import * from SMART.Java.Python.misc.Progress import * from SMART.Java.Python.misc.RPlotter import * from commons.core.parsing.ParserChooser import ParserChooser def writeCVSfile(outHandler): for pos in range(len(letters)): posTrue = pos +1 outHandler.write( "%s;" % (posTrue)) for letter in lettersRate: if positionRate[letter].has_key(pos): outHandler.write("%s=%.2f%s;" %(letter, positionRate[letter][pos], "%")) else: outHandler.write("%s=0%s;" % (letter, "%")) outHandler.write("\n") if __name__ == "__main__": # parse command line description = "Get Letter Distribution v1.0.1: Compute the distribution of nucleotides of a set of genomic coordinates. [Category: Visualization]" parser = OptionParser(description = description) parser.add_option("-i", "--input", dest="inputFileName", action="store", type="string", help="input file to be analyzed [compulsory] [format: file in sequence format given by -f]") parser.add_option("-f", "--format", dest="format", action="store", type="string", help="format of file [format: sequence file format]") parser.add_option("-o", "--output", dest="outputFileName", action="store", type="string", help="output file [compulsory] [format: output file in PNG format]") parser.add_option("-v", "--verbosity", dest="verbosity", action="store", default=1, type="int", help="trace level [format: int]") parser.add_option("-c", "--csv", dest="csv", action="store_true", default=False, help="write a .csv file [format: bool] [default: false]") parser.add_option("-l", "--log", dest="log", action="store_true", default=False, help="write a log file [format: bool] [default: false]") (options, args) = parser.parse_args() chooser = ParserChooser() chooser.findFormat(options.format) parser = chooser.getParser(options.inputFileName) nbSequences = parser.getNbSequences() print "%i sequences read" % (nbSequences) # treat items progress = Progress(nbSequences, "Analyzing sequences of " + options.inputFileName, options.verbosity) nbLettersTotal = 0 nbLetters = {} lettersRate = {} nbPositions = {} positionCount = {} positionRate = {} nbPositionRate = {} for sequence in parser.getIterator(): letters = sequence.getSequence() thisNbLettersTotal = sequence.getSize() nbLettersTotal += thisNbLettersTotal thisNbLetters = {} for pos in range(len(letters)): letter = letters[pos] if letter not in thisNbLetters: thisNbLetters[letter] = 1 else: thisNbLetters[letter] += 1 if pos+1 not in nbPositions: nbPositions[pos+1] = 1 else: nbPositions[pos+1] += 1 if letter not in positionCount: positionCount[letter] = {} if pos+1 not in positionCount[letter]: positionCount[letter][pos+1] = 1 else: positionCount[letter][pos+1] += 1 for letter in thisNbLetters: if letter not in nbLetters: nbLetters[letter] = thisNbLetters[letter] else: nbLetters[letter] += thisNbLetters[letter] if letter not in lettersRate: lettersRate[letter] = {} rate = int(float(thisNbLetters[letter]) / thisNbLettersTotal * 100) if rate not in lettersRate[letter]: lettersRate[letter][rate] = 1 else: lettersRate[letter][rate] += 1 progress.inc() progress.done() for letter in positionCount: positionRate[letter] = {} for pos in positionCount[letter]: positionRate[letter][pos] = positionCount[letter][pos] / float(nbPositions[pos]) * 100 for pos in nbPositions: nbPositionRate[pos] = nbPositions[pos] / float(nbPositions[1]) * 100 # plot content distributions plotter = RPlotter("%s.png" % (options.outputFileName), options.verbosity, True) plotter.setFill(0) plotter.setLegend(True) for letter in lettersRate: plotter.addLine(lettersRate[letter], letter) plotter.plot() # plot distribution per position plotter = RPlotter("%sPerNt.png" % (options.outputFileName), options.verbosity, True) plotter.setFill(0) plotter.setLegend(True) plotter.setXLabel("Position on the read") plotter.setYLabel("Percentage") for letter in positionRate: plotter.addLine(positionRate[letter], letter) plotter.addLine(nbPositionRate, "#") plotter.plot() if options.csv: outHandler = open("%s.csv" % (options.outputFileName), "w") writeCVSfile(outHandler) outHandler.close() print "%d sequences" % (nbSequences) print "%d letters" % (nbLettersTotal) for letter in nbLetters: print "%s: %d (%.2f%%)" % (letter, nbLetters[letter], float(nbLetters[letter]) / nbLettersTotal * 100)