def _small_file_innerdist(start,
fastq_file,
pair_file,
ref_file,
out_base,
out_dir,
config,
remove_workdir=False):
work_dir = os.path.join(out_dir, "innerdist_estimate")
if os.path.exists(work_dir):
shutil.rmtree(work_dir)
safe_makedir(work_dir)
extra_args = ["-s", str(start), "-u", "250000"]
bowtie_runner = _select_bowtie_version(config)
out_sam = bowtie_runner.align(fastq_file, pair_file, ref_file, out_base,
work_dir, config, extra_args)
dists = []
with closing(pysam.Samfile(out_sam)) as work_sam:
for read in work_sam:
if read.is_proper_pair and read.is_read1:
dists.append(abs(read.isize) - 2 * read.rlen)
if dists:
dist_stats = Stats(dists)
return int(round(dist_stats.mean())), int(
round(dist_stats.standard_deviation()))
else:
return None, None
def calc_paired_insert_stats(in_bam):
"""Retrieve statistics for paired end read insert distances.
MAD is the Median Absolute Deviation: http://en.wikipedia.org/wiki/Median_absolute_deviation
"""
dists = []
with closing(pysam.Samfile(in_bam, "rb")) as in_pysam:
for read in in_pysam:
if read.is_proper_pair and read.is_read1:
dists.append(abs(read.isize))
# remove outliers
med = Stats(dists).median()
filter_dists = filter(lambda x: x < med + 10 * med, dists)
median = Stats(filter_dists).median()
return {"mean": Stats(filter_dists).mean(), "std": Stats(filter_dists).standard_deviation(),
"median": median,
"mad": Stats([abs(x - median) for x in filter_dists]).median()}
def _estimate_paired_innerdist(fastq_file, pair_file, ref_file, out_base,
out_dir, config):
"""Use Bowtie to estimate the inner distance of paired reads.
"""
# skip initial reads for large file, but not for smaller
dists = _bowtie_for_innerdist("1000000", fastq_file, pair_file, ref_file,
out_base, out_dir, config)
if len(dists) == 0:
dists = _bowtie_for_innerdist("1", fastq_file, pair_file, ref_file,
out_base, out_dir, config, True)
dist_stats = Stats(dists)
return int(round(dist_stats.mean())), int(
round(dist_stats.standard_deviation()))
def descriptive_stats(xs):
if len(xs) < 2:
return xs
calc = Stats(xs)
parts = [
"min: %s" % min(xs),
"5%%: %s" % calc.percentile(5),
"25%%: %s" % calc.percentile(25),
"median: %s" % calc.percentile(50),
"75%%: %s" % calc.percentile(75),
"95%%: %s" % calc.percentile(95),
"99%%: %s" % calc.percentile(99),
"max: %s" % max(xs)
]
return "\n".join([" " + x for x in parts])