comparison tools/ngs_simulation/ngs_simulation.py @ 0:9071e359b9a3

Uploaded

0:9071e359b9a3

#!/usr/bin/env python

"""
Runs Ben's simulation.

usage: %prog [options]
   -i, --input=i: Input genome (FASTA format)
   -g, --genome=g: If built-in, the genome being used
   -l, --read_len=l: Read length
   -c, --avg_coverage=c: Average coverage
   -e, --error_rate=e: Error rate (0-1)
   -n, --num_sims=n: Number of simulations to run
   -p, --polymorphism=p: Frequency/ies for minor allele (comma-separate list of 0-1)
   -d, --detection_thresh=d: Detection thresholds (comma-separate list of 0-1)
   -p, --output_png=p: Plot output
   -s, --summary_out=s: Whether or not to output a file with summary of all simulations
   -m, --output_summary=m: File name for output summary of all simulations
   -f, --new_file_path=f: Directory for summary output files

"""
# removed output of all simulation results on request (not working)
#   -r, --sim_results=r: Output all tabular simulation results (number of polymorphisms times number of detection thresholds)
#   -o, --output=o: Base name for summary output for each run

from rpy import *
import os
import random, sys, tempfile
from galaxy import eggs
import pkg_resources; pkg_resources.require( "bx-python" )
from bx.cookbook import doc_optparse

def stop_err( msg ):
    sys.stderr.write( '%s\n' % msg )
    sys.exit()

def __main__():
    #Parse Command Line
    options, args = doc_optparse.parse( __doc__ )
    # validate parameters
    error = ''
    try:
        read_len = int( options.read_len )
        if read_len <= 0:
            raise Exception, ' greater than 0'
    except TypeError, e:
        error = ': %s' % str( e )
    if error:
        stop_err( 'Make sure your number of reads is an integer value%s' % error )
    error = ''
    try:
        avg_coverage = int( options.avg_coverage )
        if avg_coverage <= 0:
            raise Exception, ' greater than 0'
    except Exception, e:
        error = ': %s' % str( e )
    if error:
        stop_err( 'Make sure your average coverage is an integer value%s' % error )
    error = ''
    try:
        error_rate = float( options.error_rate )
        if error_rate >= 1.0:
            error_rate = 10 ** ( -error_rate / 10.0 )
        elif error_rate < 0:
            raise Exception, ' between 0 and 1'
    except Exception, e:
        error = ': %s' % str( e )
    if error:
        stop_err( 'Make sure the error rate is a decimal value%s or the quality score is at least 1' % error )
    try:
        num_sims = int( options.num_sims )
    except TypeError, e:
        stop_err( 'Make sure the number of simulations is an integer value: %s' % str( e ) )
    if len( options.polymorphism ) > 0:
        polymorphisms = [ float( p ) for p in options.polymorphism.split( ',' ) ]
    else:
        stop_err( 'Select at least one polymorphism value to use' )
    if len( options.detection_thresh ) > 0:
        detection_threshes = [ float( dt ) for dt in options.detection_thresh.split( ',' ) ]
    else:
        stop_err( 'Select at least one detection threshold to use' )

    # mutation dictionaries
    hp_dict = { 'A':'G', 'G':'A', 'C':'T', 'T':'C', 'N':'N' } # heteroplasmy dictionary
    mt_dict = { 'A':'C', 'C':'A', 'G':'T', 'T':'G', 'N':'N'} # misread dictionary

    # read fasta file to seq string
    all_lines = open( options.input, 'rb' ).readlines()
    seq = ''
    for line in all_lines:
        line = line.rstrip() 
        if line.startswith('>'):
            pass
        else:
            seq += line.upper()
    seq_len = len( seq )

    # output file name template
# removed output of all simulation results on request (not working)
#    if options.sim_results == "true":
#        out_name_template = os.path.join( options.new_file_path, 'primary_output%s_' + options.output + '_visible_tabular' )
#    else:
#        out_name_template = tempfile.NamedTemporaryFile().name + '_%s'
    out_name_template = tempfile.NamedTemporaryFile().name + '_%s'
    print 'out_name_template:', out_name_template

    # set up output files
    outputs = {}
    i = 1
    for p in polymorphisms:
        outputs[ p ] = {}
        for d in detection_threshes:
            outputs[ p ][ d ] = out_name_template % i
            i += 1

    # run sims
    for polymorphism in polymorphisms:
        for detection_thresh in detection_threshes:
            output = open( outputs[ polymorphism ][ detection_thresh ], 'wb' )
            output.write( 'FP\tFN\tGENOMESIZE=%s\n' % seq_len )
            sim_count = 0
            while sim_count < num_sims:
                # randomly pick heteroplasmic base index
                hbase = random.choice( range( 0, seq_len ) )
                #hbase = seq_len/2#random.randrange( 0, seq_len )
                # create 2D quasispecies list
                qspec = map( lambda x: [], [0] * seq_len )
                # simulate read indices and assign to quasispecies
                i = 0
                while i < ( avg_coverage * ( seq_len / read_len ) ): # number of reads (approximates coverage)
                    start = random.choice( range( 0, seq_len ) )
                    #start = seq_len/2#random.randrange( 0, seq_len ) # assign read start
                    if random.random() < 0.5: # positive sense read
                        end = start + read_len # assign read end
                        if end > seq_len: # overshooting origin
                            read = range( start, seq_len ) + range( 0, ( end - seq_len ) )
                        else: # regular read
                            read = range( start, end )
                    else: # negative sense read
                        end = start - read_len # assign read end
                        if end < -1: # overshooting origin
                            read = range( start, -1, -1) + range( ( seq_len - 1 ), ( seq_len + end ), -1 )
                        else: # regular read
                            read = range( start, end, -1 )
                    # assign read to quasispecies list by index
                    for j in read:
                        if j == hbase and random.random() < polymorphism: # heteroplasmic base is variant with p = het
                            ref = hp_dict[ seq[ j ] ]
                        else: # ref is the verbatim reference nucleotide (all positions)
                            ref = seq[ j ]
                        if random.random() < error_rate: # base in read is misread with p = err
                            qspec[ j ].append( mt_dict[ ref ] )
                        else: # otherwise we carry ref through to the end
                            qspec[ j ].append(ref)
                    # last but not least
                    i += 1
                bases, fpos, fneg = {}, 0, 0 # last two will be outputted to summary file later
                for i, nuc in enumerate( seq ):
                    cov = len( qspec[ i ] )
                    bases[ 'A' ] = qspec[ i ].count( 'A' )
                    bases[ 'C' ] = qspec[ i ].count( 'C' )
                    bases[ 'G' ] = qspec[ i ].count( 'G' )
                    bases[ 'T' ] = qspec[ i ].count( 'T' )
                    # calculate max NON-REF deviation
                    del bases[ nuc ]
                    maxdev = float( max( bases.values() ) ) / cov
                    # deal with non-het sites
                    if i != hbase:
                        if maxdev >= detection_thresh: # greater than detection threshold = false positive
                            fpos += 1
                    # deal with het sites
                    if i == hbase:
                        hnuc = hp_dict[ nuc ] # let's recover het variant
                        if ( float( bases[ hnuc ] ) / cov ) < detection_thresh: # less than detection threshold = false negative
                            fneg += 1
                        del bases[ hnuc ] # ignore het variant
                        maxdev = float( max( bases.values() ) ) / cov # check other non-ref bases at het site
                        if maxdev >= detection_thresh: # greater than detection threshold = false positive (possible)
                            fpos += 1
                # output error sums and genome size to summary file
                output.write( '%d\t%d\n' % ( fpos, fneg ) )
                sim_count += 1
            # close output up
            output.close()

    # Parameters (heteroplasmy, error threshold, colours)
    r( '''
    het=c(%s)
    err=c(%s)
    grade = (0:32)/32
    hues = rev(gray(grade))
    ''' % ( ','.join( [ str( p ) for p in polymorphisms ] ), ','.join( [ str( d ) for d in detection_threshes ] ) ) )

    # Suppress warnings
    r( 'options(warn=-1)' )

    # Create allsum (for FP) and allneg (for FN) objects
    r( 'allsum <- data.frame()' )
    for polymorphism in polymorphisms:
        for detection_thresh in detection_threshes:
            output = outputs[ polymorphism ][ detection_thresh ]
            cmd = '''
                  ngsum = read.delim('%s', header=T)
                  ngsum$fprate <- ngsum$FP/%s
                  ngsum$hetcol <- %s
                  ngsum$errcol <- %s
                  allsum <- rbind(allsum, ngsum)
                  ''' % ( output, seq_len, polymorphism, detection_thresh )
            r( cmd )

    if os.path.getsize( output ) == 0:
        for p in outputs.keys():
            for d in outputs[ p ].keys():
                sys.stderr.write(outputs[ p ][ d ] + ' '+str( os.path.getsize( outputs[ p ][ d ] ) )+'\n')

    if options.summary_out == "true":
        r( 'write.table(summary(ngsum), file="%s", quote=FALSE, sep="\t", row.names=FALSE)' % options.output_summary )

    # Summary objects (these could be printed)
    r( '''
    tr_pos <- tapply(allsum$fprate,list(allsum$hetcol,allsum$errcol), mean)
    tr_neg <- tapply(allsum$FN,list(allsum$hetcol,allsum$errcol), mean)
    cat('\nFalse Positive Rate Summary\n\t', file='%s', append=T, sep='\t')
    write.table(format(tr_pos, digits=4), file='%s', append=T, quote=F, sep='\t')
    cat('\nFalse Negative Rate Summary\n\t', file='%s', append=T, sep='\t')
    write.table(format(tr_neg, digits=4), file='%s', append=T, quote=F, sep='\t')
    ''' % tuple( [ options.output_summary ] * 4 ) )

    # Setup graphs
    #pdf(paste(prefix,'_jointgraph.pdf',sep=''), 15, 10)
    r( '''
    png('%s', width=800, height=500, units='px', res=250)
    layout(matrix(data=c(1,2,1,3,1,4), nrow=2, ncol=3), widths=c(4,6,2), heights=c(1,10,10))
    ''' % options.output_png )

    # Main title
    genome = ''
    if options.genome:
        genome = '%s: ' % options.genome
    r( '''
    par(mar=c(0,0,0,0))
    plot(1, type='n', axes=F, xlab='', ylab='')
    text(1,1,paste('%sVariation in False Positives and Negatives (', %s, ' simulations, coverage ', %s,')', sep=''), font=2, family='sans', cex=0.7)
    ''' % ( genome, options.num_sims, options.avg_coverage ) )

    # False positive boxplot
    r( '''
    par(mar=c(5,4,2,2), las=1, cex=0.35)
    boxplot(allsum$fprate ~ allsum$errcol, horizontal=T, ylim=rev(range(allsum$fprate)), cex.axis=0.85)
    title(main='False Positives', xlab='false positive rate', ylab='')
    ''' )

    # False negative heatmap (note zlim command!)
    num_polys = len( polymorphisms )
    num_dets = len( detection_threshes )
    r( '''
    par(mar=c(5,4,2,1), las=1, cex=0.35)
    image(1:%s, 1:%s, tr_neg, zlim=c(0,1), col=hues, xlab='', ylab='', axes=F, border=1)
    axis(1, at=1:%s, labels=rownames(tr_neg), lwd=1, cex.axis=0.85, axs='i')
    axis(2, at=1:%s, labels=colnames(tr_neg), lwd=1, cex.axis=0.85)
    title(main='False Negatives', xlab='minor allele frequency', ylab='detection threshold')
    ''' % ( num_polys, num_dets, num_polys, num_dets ) )

    # Scale alongside
    r( '''
    par(mar=c(2,2,2,3), las=1)
    image(1, grade, matrix(grade, ncol=length(grade), nrow=1), col=hues, xlab='', ylab='', xaxt='n', las=1, cex.axis=0.85)
    title(main='Key', cex=0.35)
    mtext('false negative rate', side=1, cex=0.35)
    ''' )

    # Close graphics
    r( '''
    layout(1)
    dev.off()
    ''' )

    # Tidy up
#    r( 'rm(folder,prefix,sim,cov,het,err,grade,hues,i,j,ngsum)' )

if __name__ == "__main__" : __main__()