def read_next(inf,strata):
while True:
v = inf.readline()
if not v: return False
num = int(v.rstrip().split("\t")[3])
if num < strata: continue
arr = v.split("\t")
res = Bed(arr[0],int(arr[1]),int(arr[2]))
res.set_payload(int(num))
return res
def get_junctions(sams,args):
prog = re.compile('^[MDNX=]$')
outsams = {}
z = 0
outs = []
for sam in sams:
z+=1
outsams[z] = sam
v = [x for x in sam.value('cigar_array') if prog.match(x['op'])]
juncs = [i for i in range(0,len(v)) if v[i]['op'] =='N' and v[i]['val'] >= args.minimum_intron_size]
for i in juncs:
coord1 = sum([x['val'] for x in v[0:i]]) + sam.value('pos')
coord2 = coord1 + v[i]['val']
b1 = Bed(sam.value('rname'),coord1-2,coord1-1)
b2 = Bed(sam.value('rname'),coord2-1,coord2)
outs.append([Junction(b1,b2),z])
return [outs,outsams]
def main():
#do our inputs
args = do_inputs()
sys.stderr.write("Reading reference genepred\n")
ref = {}
tx_strand = {}
z = 0
with open(args.reference_genepred) as inf:
for line in inf:
gpd = GPD(line)
gname = gpd.get_gene_name()
tname = gpd.get_transcript_name()
tx_strand[tname] = gpd.get_strand()
if gname not in ref: ref[gname] = []
ref[gname].append(gpd)
z += 1
sys.stderr.write("Read "+str(len(ref.keys()))+" genes and "+str(z)+" transcripts\n")
if args.maximum_isoforms > 0:
sys.stderr.write("Removing genes with more than "+str(args.maximum_isoforms)+" isoforms.\n")
for gname in ref.keys():
if len(ref[gname]) > args.maximum_isoforms: del ref[gname]
sys.stderr.write("Now have "+str(len(ref.keys()))+" genes and "+str(sum([len(ref[x]) for x in ref.keys()]))+" transcripts\n")
sys.stderr.write("Filtering by length "+str(args.minimum_length)+" bp\n")
for gname in ref.keys():
passing = []
for gpd in ref[gname]:
if gpd.get_length() < args.minimum_length: continue
passing.append(gpd)
if len(passing) == 0: del ref[gname]
else: ref[gname] = passing
sys.stderr.write("Now have "+str(len(ref.keys()))+" genes and "+str(sum([len(ref[x]) for x in ref.keys()]))+" transcripts\n")
sys.stderr.write("Converting gpd into exon bed\n")
beds = []
for gname in ref.keys():
for gpd in ref[gname]:
tname = gpd.get_transcript_name()
for i in range(0,len(gpd.exons)):
ex = gpd.exons[i]
beds.append(ex.get_range().get_bed_array()+[gname,tname,i])
with open(args.tempdir+'/gpd.bed','w') as of:
for bed in sorted(beds,key=lambda x: (x[0],x[1],x[2],x[3],x[4],x[5])):
of.write("\t".join([str(x) for x in bed])+"\n")
sys.stderr.write("intersecting with bed depth\n")
of = open(args.tempdir+'/intersect.bed','w')
cmd = 'bedtools intersect -wo -a - -b '+args.tempdir+'/gpd.bed'
p = Popen(cmd.split(),stdin=args.bed_depth,stdout=of)
p.communicate()
coverage = {}
sys.stderr.write("Reading the intersection\n")
with open(args.tempdir+'/intersect.bed') as inf:
for line in inf:
f = line.rstrip().split("\t")
gname = f[7]
tname = f[8]
depth = int(f[3])
bed1 = Bed(f[0],int(f[1]),int(f[2]))
bed2 = Bed(f[4],int(f[5]),int(f[6]))
bed = bed1.union(bed2)
bed.set_payload(depth)
if gname not in coverage:
coverage[gname] = {}
if tname not in coverage[gname]:
coverage[gname][tname] = []
coverage[gname][tname].append(bed)
transcript_depths = {}
for gname in coverage:
for tname in coverage[gname]:
ref_gpd = [x for x in ref[gname] if x.get_transcript_name()==tname][0]
rlen = ref_gpd.get_length()
bases_covered = sum([x.length() for x in coverage[gname][tname]])
bases_area = sum([x.length()*x.get_payload() for x in coverage[gname][tname]])
avg_depth = float(bases_area)/float(rlen)
if avg_depth < args.minimum_average_depth: continue
if bases_covered < args.minimum_length: continue
#print gname
#print tname
#print rlen
#print bases_covered
#print bases_area
total_positions = {}
for ex in ref_gpd.exons:
b = ex.get_range().get_bed_array()
for i in range(b[1],b[2]):
total_positions[i] = 0 # zero indexed
for b in coverage[gname][tname]:
depth = b.get_payload()
barr = b.get_bed_array()
for i in range(barr[1],barr[2]):
total_positions[i] = depth
transcript_depths[tname] = total_positions
sys.stderr.write("have information needed to plot from "+str(len(transcript_depths.keys()))+" transcripts\n")
outputs = []
for tname in transcript_depths:
depths = transcript_depths[tname]
positions = sorted(depths.keys())
tx_len = len(positions)
bins = {}
for i in range(0,tx_len):
bin = int(100*float(i)/float(tx_len))
if bin not in bins: bins[bin] = []
bins[bin].append(depths[positions[i]])
for bin in bins:
bins[bin] = average(bins[bin])
biggest = float(max(bins.values()))
tx_array = [float(bins[x])/biggest for x in sorted(bins.keys())]
if tx_strand[tname] == '-':
tx_array.reverse()
#outputs.append(tx_array)
args.output.write(tname+"\t"+"\t".join([str(x) for x in tx_array])+"\n")
#for i in range(0,100):
# args.output.write("\t".join([str(x[i]) for x in outputs])+"\n")
args.output.close()
# Temporary working directory step 3 of 3 - Cleanup
if not args.specific_tempdir:
rmtree(args.tempdir)
def get_overlap(fileA,fileB,min_A,min_B):
infA = do_open(fileA)
infB = do_open(fileB)
bufA = read_next(infA,min_A)
bufB = read_next(infB,min_B)
tot = 0
sizeA = 0
sizeB = 0
if bufA:
sizeA = bufA.length()
if bufB:
sizeB = bufB.length()
zA = 1
zB = 1
while True:
#if (zA%10000 ==0 or zB%10000==0): sys.stderr.write(str(zA)+" "+str(zB)+" \r")
if not bufA or not bufB: break
c = bufA.cmp(bufB)
if c == 0:
tot += bufA.overlap_size(bufB)
saveA = bufA
nA = bufA.subtract(bufB)
if len(nA) > 0 and nA[-1].end == bufA.end:
num = bufA.get_payload()
bufA = Bed(nA[-1].chr,nA[-1].start-1,nA[-1].end)
bufA.set_payload(num)
else:
bufA = read_next(infA,min_A)
if bufA:
sizeA += bufA.length()
zA+=1
nB = bufB.subtract(saveA)
if len(nB) > 0 and nB[-1].end == bufB.end:
num = bufB.get_payload()
bufB = Bed(nB[-1].chr,nB[-1].start-1,nB[-1].end)
bufB.set_payload(num)
else:
bufB = read_next(infB,min_B)
if bufB:
sizeB += bufB.length()
zB+=1
elif c == -1:
bufA = read_next(infA,min_A)
if bufA:
sizeA += bufA.length()
zA += 1
else:
bufB = read_next(infB,min_B)
if bufB:
sizeB += bufB.length()
zB += 1
#sys.stderr.write("\n")
if bufA:
while True:
bufA = read_next(infA,min_A)
if bufA: sizeA += bufA.length()
else: break
if bufB:
while True:
bufB = read_next(infB,min_B)
if bufB: sizeB += bufB.length()
else: break
infA.close()
infB.close()
return [min_A,min_B,sizeA,sizeB,tot]
def main():
parser = argparse.ArgumentParser(
description="", formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('input', help="Use - for STDIN")
parser.add_argument('--offset',
type=int,
default=0,
help="add this much to transcript tpms")
parser.add_argument('--mult',
type=int,
default=10,
help="multiply this much to tpms")
parser.add_argument('--min_exons',
type=int,
default=1,
help="require at least this many exons")
parser.add_argument('-o',
'--output',
help="OUTPUT file or nothing for STDOUT")
args = parser.parse_args()
inf = sys.stdin
if args.input != '-':
if args.input[-3:] == '.gz':
inf = gzip.open(args.input)
else:
inf = open(args.input)
genes = {}
sys.stderr.write("Reading gtf file\n")
txs = {}
for line in inf:
if re.match('#', line): continue
f = line.rstrip().split("\t")
tx = None
if f[2] == 'exon' or f[2] == 'transcript':
tx = re.search('transcript_id\s+"([^"]+)"', f[8]).group(1)
if tx not in txs:
txs[tx] = {}
txs[tx]['tpm'] = 0
txs[tx]['exons'] = []
if f[2] == 'transcript':
tpm = float(re.search('TPM\s+"([^"]+)"', f[8]).group(1))
txs[tx]['tpm'] = int((tpm * float(args.mult)) + args.offset)
if f[2] == 'exon':
chr = f[0]
start = int(f[3]) - 1
end = int(f[4])
txs[tx]['exons'].append(Bed(chr, start, end))
inf.close()
vals = []
sys.stderr.write("Traversing annotation file\n")
for tx in txs:
exons = txs[tx]['exons']
v = txs[tx]['tpm']
if len(exons) < args.min_exons: continue
for i in range(0, v):
vals += exons[:]
sys.stderr.write("Generating coverage file " + str(len(vals)) + "\n")
of = sys.stdout
if args.output:
if args.output[-3:] == '.gz':
of = gzip.open(args.output, 'w')
else:
of = open(args.output, 'w')
covs = ranges_to_coverage(vals)
for v in covs:
of.write(v.chr + "\t" + str(v.start - 1) + "\t" + str(v.end) + "\t" +
str(v.get_payload()) + "\n")
# of.write(tx+"\t"+gene+"\t"+str(genes[gene]['transcripts'][tx])+"\t"+str(genes[gene]['cnt'])+"\n")
of.close()
