annotate alignr.py @ 23:4e646baac551

Uploaded

'''
the scripts takes two files as input, and compute the coverage of 
features in input 1 across features in input 2. Features in input 2 are 
divided into bins and coverage is computed for each bin.  

please check the help information by typing:

    python align.py -h


requirement:
    please install the following tools first:
    bedtools:   for read/region overlapping, http://code.google.com/p/bedtools/
    
'''

import os,sys,os.path
from optparse import OptionParser

def lineCount(filename):
    with open(filename) as f:
        for i, l in enumerate(f):
            pass
    return i + 1

def combineFilename(f1,f2):
    '''
    fuse two file names into one
    '''
    return f1.split('/')[-1]+'-'+f2.split('/')[-1]

def checkFormat(filename1,filename2,input1format):
    '''
    check the format of input files
    '''

    # file1
    # read the first line, see how many filds
    ncol1 = 6
    if input1format == "BED":
        f = open(filename1)
        line = f.readline().strip().split('\t')
        ncol1 = len(line)
        if ncol1 < 3:
            print "ERROR: "+filename1+" has only "+str(ncol1)+" columns (>=3 required). Make sure it has NO header line and is TAB-delimited."
            sys.exit(1)
        f.close()
     
    # file2
    f = open(filename2)
    line = f.readline().strip().split('\t')
    ncol2 = len(line)  
    if ncol2 < 3:
        print "ERROR: "+filename2+" has only "+str(ncol2)+" columns (>=3 required). Make sure it has NO header line and is TAB-delimited."
        sys.exit(1)        

    return ncol1,ncol2


def makeBed(filename,ncol):
    '''
    add up to 6 column
    '''
    f = open(filename)
    outfile = filename+'.tmp.bed'
    outf = open(outfile,'w')
    if ncol == 3:
        for line in f:
            outf.write(line.strip()+'\t.\t0\t+\n')
    elif ncol == 4:
        for line in f:
            outf.write(line.strip()+'\t0\t+\n')
    if ncol == 5:
        for line in f:
            outf.write(line.strip()+'\t+\n')
    f.close()
    outf.close()
    return outfile
    
def makeWindow(filename,window):

    outfile = filename+'-window='+str(window)+'.tmp.bed'
    if not os.path.exists(outfile):
        f=open(filename)
        out = open(outfile,'w')
        lines = f.readlines()
        if 'track' in lines[0]:
            del lines[0]
        for line in lines:
            flds = line.strip().split('\t')

            #new position
            center = (int(flds[1]) + int(flds[2]))/2
            start = center - window
            end = center + window
            if start >= 0:
                flds[1] = str(start)
                flds[2] = str(end)
                out.write('\t'.join(flds)+'\n')
        f.close()
        out.close()
    return outfile

def groupReadsMapped2aRegion(filename,ncol):
    '''
    read output from intersectBED
    find all reads mapped to each region
    '''
    try:
        f=open(filename)
        #If filename cannot be opened, print an error message and exit
    except IOError:
        print "could not open",filename,"Are you sure this file exists?"
        sys.exit(1)
    lines = f.readlines()
    
    allReadsStart = {}
    allReadsEnd = {}
    regionStrand = {}
    regionStart = {}
    regionEnd = {}
    
    for line in lines:
        flds = line.strip().split('\t')
        key = '_'.join(flds[ncol:(ncol+4)])
        if not allReadsStart.has_key(key):
            allReadsStart[key] = list()
            allReadsEnd[key] = list()
        #print flds[ncol+0],flds[ncol+1],flds[ncol+2]
        allReadsStart[key].append(int(flds[1]))
        allReadsEnd[key].append(int(flds[2]))
        regionStrand[key] = flds[ncol+5]  
        regionStart[key] = int(flds[ncol+1])    
        regionEnd[key] = int(flds[ncol+2])      
    return (allReadsStart,allReadsEnd,regionStrand,regionStart,regionEnd)
            
            
def createRegionProfile(allReadsStart,allReadsEnd,regionStrand,regionStart,regionEnd,nbins):
    '''
    each region is divided into nbins
    compute the number of reads covering each bin for each region 
    '''
    RegionProfile = {}
    nRead = {}  # num of all reads in the region
    for region in allReadsStart.keys():
        RegionProfile[region] = [0]*nbins
        nRead[region] = len(allReadsStart[region])
        #print region,nRead[region],allReadsStart[region]
        for i in range(nRead[region]):
            RegionProfile[region] = updateRegionCount(RegionProfile[region],allReadsStart[region][i],allReadsEnd[region][i],regionStart[region],regionEnd[region],regionStrand[region],nbins)
    return RegionProfile,nRead
    
def updateRegionCount(RegionCount,readStart,readEnd,regionStart,regionEnd,strand,nbins):
    '''
    each region is divided into nbins,
    add 1 to each bin covered by the read  
    '''
    L = regionEnd-regionStart
    start = int(nbins*(readStart-regionStart)/L)
    end = int(nbins*(readEnd-regionStart)/L)
    if start < 0:
        start = 0
    if end > nbins:
        end = nbins
    if strand == '-':        
        for i in range(start,end):
            RegionCount[nbins-1-i] = RegionCount[nbins-1-i] + 1
    else: # if the 6th column of the input is not strand, will treat as + strand by default       
        for i in range(start,end):
            RegionCount[i] = RegionCount[i] + 1            
    return RegionCount

def saveProfile(filename,Profile,nRegion):
    out = open(filename,'w')
    for regionType in Profile.keys():
        #print Profile[regionType]
        out.write(regionType+'\t'+str(nRegion[regionType])+'\t'+'\t'.join(map(str,Profile[regionType]))+'\n')    
                    
def saveSummary(filename,Profile,nbin):
    out = open(filename+'.summary','w')

    nfeat = len(Profile)
    summaryprofile = [0]*nbin
    for regionType in Profile.keys():
        for i in range(nbin):
            summaryprofile[i] += Profile[regionType][i]    
    out.write(filename+'\t'+str(nfeat)+'\t'+'\t'.join(map(str,summaryprofile))+'\n')  
    out.close()
    # calculate standard error
    out = open(filename+'.standarderror','w')
    sd = [0.0]*nbin
    u = [0.0]*nbin 
    for i in range(nbin):
        u[i] = float(summaryprofile[i])/nfeat
        for regionType in Profile.keys():
            sd[i] = sd[i] + (Profile[regionType][i] - u[i])**2
        sd[i] = sd[i]**0.5 / nfeat
    out.write(filename+'\t'+str(nfeat)+'\t'+'\t'.join(map(str,sd))+'\n')  
    out.close()    
                
def main():
    usage = "usage: %prog [options] -a inputA -b inputB"
    parser = OptionParser(usage)
    parser.add_option("-a", dest="inputA",
                      help="(required) input file A, interval (first 3 columns are chrN, start and end) or BAM format. The script computes the depth of coverage of features in file A across the features in file B" )                                                
    parser.add_option("-b",dest="inputB",
                      help="(required) input file B, interval file" )                                                
    parser.add_option("-f",dest="aformat",default="BED",
                      help="Format of input file A. Can be BED (default) or BAM")
    parser.add_option("-w",type='int',dest="window",
                      help="Generate new inputB by making a window of 2 x WINDOW bp (in total) flanking the center of each input feature" )     
    parser.add_option("-n", type="int", dest="nbins",default=100,
                        help="number of bins. Features in B are binned, and the coverage is computed for each bin. Default is 100")                    
    parser.add_option("-s",action="store_true", dest="strandness",
                      help="enforce strandness: require overlapping on the same strand. Default is off")
    parser.add_option("-p",action="store_true", dest="plot",default=False,
                      help="load existed intersectBed outputfile")
    parser.add_option("-q", action="store_true", dest="quiet",default=False,
                        help="suppress output on screen")
    parser.add_option("-o", dest="output_data",
                      help="(optional) output coverage file (txt) name." )
    parser.add_option("-v", dest="output_plot",
                      help="(optional) output plot (pdf) file name." )
    parser.add_option("-l", dest="plot_title", default="",
                      help="(optional) output title of the plot." )
    parser.add_option("--ylim", dest="ylim", default="min,max",
                      help="(optional) ylim of the plot" )
    parser.add_option("--summary-only", action="store_true", dest="summary_only",default=False,
                        help="save profile summary only (no data for individual features)")
    parser.add_option("--compute-se", action="store_true", dest="compute_se",default=False,
                        help="compute and plot standard deviation for each bin. used when --summary-only is on")
    parser.add_option("--profile-only", action="store_true", dest="profile_only",default=False,
                        help="save profile only (no plot)")
    parser.add_option("--span", type="float", dest="span",default=0.1,
                        help="loess span smooth parameter, 0.1 ~ 1")                    
    
    (options, args) = parser.parse_args()

    if options.inputA == None or options.inputB == None:
        parser.error("Please specify two input files!!")

    if not options.quiet:
        print "checking input file format..."
        
    ncol,ncol2 = checkFormat(options.inputA ,options.inputB,options.aformat)

    if ncol2 < 6:
        options.inputB = makeBed(options.inputB,ncol2)        
        if not options.quiet:
            print "fill up 6 columns"

    if options.window > 0:
        if not options.quiet:
            print "making windows from "+options.inputB+"..." 
        options.inputB = makeWindow(options.inputB,options.window)
    
    output = combineFilename(str(options.inputA),str(options.inputB))
    
    if not options.plot:
        if options.aformat == "BAM":
            cmd = "intersectBed -abam "+str(options.inputA)+" -b "+str(options.inputB) + ' -bed -split '
        else:
            cmd = "intersectBed -a "+str(options.inputA)+" -b "+str(options.inputB)
        if options.strandness:
            cmd = cmd + ' -s'
        cmd = cmd +" -wo > "+ output+'-intersect.tmp.bed'
        if not options.quiet:
            print "search for overlappings: "+cmd
        status = os.system(cmd)
        if status != 0:
            sys.exit(1)

    
    if not options.quiet:
        print 'group reads mapped to the same region...'
    
    allReadsStart,allReadsEnd,regionStrand,regionStart,regionEnd = groupReadsMapped2aRegion(output+'-intersect.tmp.bed',ncol)

    if len(allReadsStart) == 0:
        if not options.quiet:
            print 'no overlap found!!'
        os.system('rm *tmp.*')
        sys.exit(1)
    
    if not options.quiet:
        print 'count number of reads mapped to each bin...'
    
    RegionProfile,nRead = createRegionProfile(allReadsStart,allReadsEnd,regionStrand,regionStart,regionEnd,options.nbins) 
   
    if options.output_data == None:
        options.output_data = output+'.txt'

    if options.summary_only:  
        saveSummary(options.output_data,RegionProfile,options.nbins) 
    
    else:                 
        saveProfile(options.output_data,RegionProfile,nRead)
    
    if not options.quiet:
        print 'results saved to: '+ options.output_data 
        
    if not (options.summary_only or options.profile_only ):          
        # visualize 

        if options.window < 1:
            xlab = 'relative position (bins)'
        else:
            xlab = 'relative position (bp)'
	            
        if options.output_plot == None:
            options.output_plot = output+'.pdf'

        title = options.plot_title+'\n n = '+str(len(RegionProfile))

        rscript = open("tmp.r","w")
        rscript.write("x <- read.table('"+options.output_data+"')\n")
        rscript.write("pdf('"+options.output_plot+"')\n")
        rscript.write("avg <- colSums(x[,3:ncol(x)])/nrow(x)\n")
        rscript.write("err <- sd(x[,3:ncol(x)])/sqrt(nrow(x))\n")
        
        if options.window == 0:
            rscript.write("xticks <- seq("+str(options.nbins)+")\n")
        else:
            rscript.write("xticks <- seq("+str(-options.window)+","+str(options.window)+",length.out="+str(options.nbins)+")\n")

        if options.ylim != 'min,max':
            rscript.write("ylim=c("+options.ylim+")\n")
        else:
            rscript.write("ylim=c(min(avg-err),max(avg+err))\n")
        rscript.write("par(cex=1.5)\n")
        #smooth
        if options.span >= 0.1:
            rscript.write("avg = loess(avg~xticks,span="+str(options.span)+")$fitted\n")
            rscript.write("err = loess(err~xticks,span="+str(options.span)+")$fitted\n")
        rscript.write("plot(xticks,avg,ylab='average coverage',main='"+title+"',xlab='"+xlab+"',type='l',lwd=0,ylim=ylim)\n")   
        rscript.write("polygon(c(xticks,rev(xticks)),c(avg+err,rev(avg-err)),col='slateblue1',border=NA)\n")
        rscript.write("lines(xticks,avg,type='l',lwd=1)\n")   
        #rscript.write("xticks <- barplot(avg,names.arg=seq("+str(options.nbins)+"),ylab='average coverage',main='"+title+"',xlab='"+xlab+"',,ylim=c(min(avg-err),max(avg+err)))\n")
        #rscript.write("arrows(xticks,avg+err, xticks, avg-err, angle=90, code=3, length=0.0,col='green')\n")
        #rscript.write("lines(xticks,avg,lwd=2)\n")
        #rscript.write("lines(xticks,avg-err,col='green')\n")
        #rscript.write("lines(xticks,avg+err,col='green')\n")
        rscript.write("dev.off()\n")
        rscript.close()

        os.system("R --vanilla < tmp.r")    
    
    # remove intermediate output
    os.system('rm *tmp.*')

    
if __name__ == "__main__":
    main()