def make_read_length_file(fastq, out_dir):
basename = os.path.basename(strip_ext_fastq(fastq))
prefix = os.path.join(out_dir, basename)
txt = '{}.read_length.txt'.format(prefix)
read_length = get_read_length(fastq)
with open(txt, 'w') as fp:
fp.write(str(read_length))
return txt
def bwa_pe(fastq1, fastq2, ref_index_prefix, nth, use_bwa_mem_for_pe, out_dir):
basename = os.path.basename(strip_ext_fastq(fastq1))
prefix = os.path.join(out_dir, basename)
sam = '{}.sam'.format(prefix)
badcigar = '{}.badReads'.format(prefix)
bam = '{}.bam'.format(prefix)
temp_files = []
read_len = get_read_length(fastq1)
if use_bwa_mem_for_pe and read_len >= 70:
cmd = 'bwa mem -M -t {} {} {} {} | gzip -nc > {}'
cmd = cmd.format(nth, ref_index_prefix, fastq1, fastq2, sam)
temp_files.append(sam)
else:
sai1 = bwa_aln(fastq1, ref_index_prefix, nth, out_dir)
sai2 = bwa_aln(fastq2, ref_index_prefix, nth, out_dir)
cmd = 'bwa sampe {} {} {} {} {} | gzip -nc > {}'.format(
ref_index_prefix, sai1, sai2, fastq1, fastq2, sam)
temp_files.extend([sai1, sai2, sam])
run_shell_cmd(cmd)
cmd2 = 'zcat -f {} | '
cmd2 += 'awk \'BEGIN {{FS="\\t" ; OFS="\\t"}} ! /^@/ && $6!="*" '
cmd2 += '{{ cigar=$6; gsub("[0-9]+D","",cigar); '
cmd2 += 'n = split(cigar,vals,"[A-Z]"); s = 0; '
cmd2 += 'for (i=1;i<=n;i++) s=s+vals[i]; seqlen=length($10); '
cmd2 += 'if (s!=seqlen) print $1"\\t"; }}\' | '
cmd2 += 'sort | uniq > {}'
cmd2 = cmd2.format(sam, badcigar)
run_shell_cmd(cmd2)
# Remove bad CIGAR read pairs
if get_num_lines(badcigar) > 0:
cmd3 = 'zcat -f {} | grep -v -F -f {} | '
cmd3 += 'samtools view -Su - | samtools sort - -o {} -T {}'
cmd3 = cmd3.format(sam, badcigar, bam, prefix)
else:
cmd3 = 'samtools view -Su {} | samtools sort - -o {} -T {}'
cmd3 = cmd3.format(sam, bam, prefix)
run_shell_cmd(cmd3)
rm_f(temp_files)
return bam
def bwa_pe(fastq1, fastq2, ref_index_prefix, nth, mem_gb, use_bwa_mem_for_pe,
bwa_mem_read_len_limit, rescue_reads_for_bwa_mem, out_dir):
basename = os.path.basename(strip_ext_fastq(fastq1))
prefix = os.path.join(out_dir, basename)
sam = '{}.sam'.format(prefix)
badcigar = '{}.badReads'.format(prefix)
bam = '{}.bam'.format(prefix)
temp_files = []
read_len = get_read_length(fastq1)
log.info('Guessed read length of R1 FASTQ: {read_len}'.format(
read_len=read_len, ))
if use_bwa_mem_for_pe and read_len >= bwa_mem_read_len_limit:
log.info('Use bwa mem.')
cmd = 'bwa mem -M {extra_param} -t {nth} {ref_index_prefix} {fastq1} {fastq2} | gzip -nc > {sam}'.format(
extra_param='-P' if rescue_reads_for_bwa_mem else '',
nth=nth,
ref_index_prefix=ref_index_prefix,
fastq1=fastq1,
fastq2=fastq2,
sam=sam,
)
temp_files.append(sam)
else:
log.info('Use bwa aln for each (R1 and R2) and then bwa sampe.')
sai1 = bwa_aln(fastq1, ref_index_prefix, nth, out_dir)
sai2 = bwa_aln(fastq2, ref_index_prefix, nth, out_dir)
cmd = 'bwa sampe {} {} {} {} {} | gzip -nc > {}'.format(
ref_index_prefix, sai1, sai2, fastq1, fastq2, sam)
temp_files.extend([sai1, sai2, sam])
run_shell_cmd(cmd)
cmd2 = 'zcat -f {} | '
cmd2 += 'awk \'BEGIN {{FS="\\t" ; OFS="\\t"}} ! /^@/ && $6!="*" '
cmd2 += '{{ cigar=$6; gsub("[0-9]+D","",cigar); '
cmd2 += 'n = split(cigar,vals,"[A-Z]"); s = 0; '
cmd2 += 'for (i=1;i<=n;i++) s=s+vals[i]; seqlen=length($10); '
cmd2 += 'if (s!=seqlen) print $1"\\t"; }}\' | '
cmd2 += 'sort | uniq > {}'
cmd2 = cmd2.format(sam, badcigar)
run_shell_cmd(cmd2)
# Remove bad CIGAR read pairs
if get_num_lines(badcigar) > 0:
run_shell_cmd(
'zcat -f {sam} | grep -v -F -f {badcigar} | '
'samtools view -Su /dev/stdin | samtools sort /dev/stdin -o {bam} -T {prefix} {res_param}'
.format(
sam=sam,
badcigar=badcigar,
bam=bam,
prefix=prefix,
res_param=get_samtools_res_param('sort',
nth=nth,
mem_gb=mem_gb),
))
else:
run_shell_cmd(
'samtools view -Su {sam} | samtools sort /dev/stdin -o {bam} -T {prefix} {res_param}'
.format(
sam=sam,
bam=bam,
prefix=prefix,
res_param=get_samtools_res_param('sort',
nth=nth,
mem_gb=mem_gb),
))
rm_f(temp_files)
return bam