def buildCodingPotential(infile, outfile):
'''run CPC analysis as in the cpc script.
This module runs framefinder and blastx on both strands.
It seems to work, but I have not thoroughly tested it.
I expect that the false positive rate increases (i.e.,
predicting non-coding as coding) in cases where the best
framefinder match and the best blast match are on opposite
strands. In the original CPC, these would be separated.
'''
try:
cpc_dir = os.environ["CPC_HOME"]
except KeyError:
raise ValueError("CPC_HOME environment variable is not set. ")
tmpdir = P.getTempDir(".")
track = P.snip(outfile, ".coding.gz")
# extract features for frame finder
# replaces extract_framefinder_feats.pl to parse both strands
with open(os.path.join(tmpdir, "ff.feat"), "w") as outf:
outf.write(
"\t".join(("QueryID", "CDSLength", "Score", "Used", "Strict")) + "\n")
for line in IOTools.openFile("%s.frame.gz" % track):
if line.startswith(">"):
try:
(id, start, end, score, used, mode, tpe) = \
re.match(
">(\S+).*framefinder \((\d+),(\d+)\) score=(\S+) used=(\S+)% \{(\S+),(\w+)\}", line).groups()
except AttributeError:
raise ValueError("parsing error in line %s" % line)
length = int(end) - int(start) + 1
strict = int(tpe == "strict")
outf.write(
"\t".join((id, str(length), used, str(strict))) + "\n")
to_cluster = USECLUSTER
# extract features and prepare svm data
s = []
s.append('''
zcat %(infile)s
| perl %(cpc_dir)s/libs/blast2table.pl
| tee %(tmpdir)s/blastx.table
| perl %(cpc_dir)s/bin/extract_blastx_features.pl
> %(tmpdir)s/blastx.feat1;
''')
s.append('''
cat %(track)s_norepeats.fasta
| perl %(cpc_dir)s/bin/add_missing_entries.pl
%(tmpdir)s/blastx.feat1
> %(tmpdir)s/blastx.feat;
''')
# step 2 - prepare data
s.append('''
perl %(cpc_dir)s/bin/feat2libsvm.pl -c 2,4,6 NA NA %(tmpdir)s/blastx.feat
> %(tmpdir)s/blastx.lsv;
''')
s.append('''
perl %(cpc_dir)s/bin/feat2libsvm.pl -c 2,3,4,5 NA NA %(tmpdir)s/ff.feat
> %(tmpdir)s/ff.lsv;
''')
s.append('''
perl -w %(cpc_dir)s/bin/lsv_cbind.pl %(tmpdir)s/blastx.lsv %(tmpdir)s/ff.lsv
> %(tmpdir)s/test.lsv;
''')
s.append('''
%(cpc_dir)s/libs/libsvm/libsvm-2.81/svm-scale
-r %(cpc_dir)s/data/libsvm.range
%(tmpdir)s/test.lsv
> %(tmpdir)s/test.lsv.scaled;
''')
# step 3: prediction
m_libsvm_model0 = os.path.join(cpc_dir, "data/libsvm.model0") # standard
m_libsvm_model = os.path.join(cpc_dir, "data/libsvm.model") # Prob
m_libsvm_model2 = os.path.join(
cpc_dir, "data/libsvm.model2") # Prob + weighted version
m_libsvm_range = os.path.join(cpc_dir, "data/libsvm.range")
s.append('''
%(cpc_dir)s/libs/libsvm/libsvm-2.81/svm-predict2
%(tmpdir)s/test.lsv.scaled
%(m_libsvm_model0)s
%(tmpdir)s/test.svm0.predict
> %(tmpdir)s/test.svm0.stdout 2> %(tmpdir)s/test.svm0.stderr;
''')
s.append('''
printf "gene_id\\tlength\\tresult\\tvalue\\n"
| gzip > %(outfile)s;
cat %(tmpdir)s/test.svm0.predict
| perl -w %(cpc_dir)s/bin/predict.pl %(track)s_norepeats.fasta
| gzip >> %(outfile)s;
''')
# generate reports
s.append('''cat %(tmpdir)s/blastx.feat
| perl -w %(cpc_dir)s/bin/generate_plot_features.pl %(tmpdir)s/blastx.table <( zcat %(track)s.frame.gz)
| perl -w %(cpc_dir)s/bin/split_plot_features_by_type.pl %(outfile)s.homology %(outfile)s.orf;
gzip %(outfile)s.orf %(outfile)s.homology;
''')
# now run it all
statement = " checkpoint; ".join(s)
P.run()
# clean up
shutil.rmtree(tmpdir)
def filterBamfiles(infile, sentinel):
"""
Pre-process bamfiles prior to peak calling.
i) sort bamfiles
ii) remove unmapped readswith bam2bam.py
iii) remove non-uniquely mapping reads with bam2bam.py (optional)
iv) remove duplicates with Picards MarkDuplicates (optional)
v) remove reads from masked regions with bedtools intersect (optional)
vi) index
"""
# create tempfile for Picard's MarkDuplicates
picard_tmp = P.getTempDir(PARAMS["scratchdir"])
outfile = P.snip(sentinel, ".sentinel") + ".bam"
# ensure bamfile is sorted,
statement = [
"samtools sort @IN@ -o @[email protected]",
]
# remove unmapped reads
statement.append("cgat bam2bam"
" --method=filter --filter-method=mapped"
" --log=%(outfile)s.log"
" < @[email protected]"
" > @OUT@")
# remove non-uniquely mapping reads, if requested
if PARAMS["filter_remove_non_unique"]:
statement.append("cgat bam2bam"
" --method=filter --filter-method=unique"
" --log=%(outfile)s.log"
" < @IN@"
" > @OUT@")
# remove duplicates, if requested
if PARAMS["filter_remove_duplicates"]:
statement.append("MarkDuplicates"
" INPUT=@IN@"
" ASSUME_SORTED=true"
" REMOVE_DUPLICATES=true"
" QUIET=false"
" OUTPUT=@OUT@"
" METRICS_FILE=/dev/null"
" VALIDATION_STRINGENCY=SILENT"
" TMP_DIR=%(picard_tmp)s"
" 2> %(outfile)s.log")
# mask regions, if intervals supplied
if PARAMS["filter_mask_intervals"]:
mask = PARAMS["filter_mask_intervals"]
statement.append("bedtools intersect"
" -abam @IN@"
" -b %(mask)s"
" -wa"
" -v"
" > @OUT@")
statement.append("mv @IN@ %(outfile)s")
statement.append("samtools index %(outfile)s")
job_memory = "5G"
statement = P.joinStatements(statement, infile)
P.run()
P.touch(sentinel)
shutil.rmtree(picard_tmp)
