def get_loci(transcripts_genepred):
loci = Loci()
loci.verbose = True
with open(transcripts_genepred) as inf:
for line in inf:
if line[0] == '#': continue
gpd = GenePredEntry(line.rstrip())
rng = Bed(gpd.value('chrom'), gpd.value('txStart'),
gpd.value('txEnd'))
rng.set_payload(gpd.value('name'))
loc1 = Locus()
loc1.add_member(rng)
loci.add_locus(loc1)
sys.stderr.write("Organizing genepred data into overlapping loci\n")
sys.stderr.write("Started with " + str(len(loci.loci)) + " loci\n")
loci.update_loci()
sys.stderr.write("Ended with " + str(len(loci.loci)) + " loci\n")
m = 0
locus2name = {}
name2locus = {}
for locus in loci.loci:
m += 1
for member in locus.members:
name = member.get_payload()
if m not in locus2name: locus2name[m] = set()
locus2name[m].add(name)
name2locus[name] = m
return [locus2name, name2locus]
def read_first(self, ingpd):
self.gpds.append(ingpd)
sjun = get_simple_junction(ingpd)
if sjun:
self.simple_junction_set.add(sjun)
if self.params["use_dir"]:
self.dir = ingpd.value("strand")
# add fuzzy junctions
chr = ingpd.value("chrom")
for i in range(0, len(ingpd.value("exonStarts")) - 1):
self.fuzzy_junctions.append(
FuzzyJunction(chr, ingpd.value("exonEnds")[i], ingpd.value("exonStarts")[i + 1] + 1, self.dir)
)
if len(ingpd.value("exonStarts")) > 1: # we have junctions
self.fuzzy_junctions[0].left.get_payload()["start"] = Bed(
chr, ingpd.value("txStart"), ingpd.value("txStart") + 1, self.dir
)
self.fuzzy_junctions[0].left.get_payload()["start"].set_payload([])
self.fuzzy_junctions[0].left.get_payload()["start"].get_payload().append(ingpd.value("txStart") + 1)
self.fuzzy_junctions[-1].right.get_payload()["end"] = Bed(
chr, ingpd.value("txEnd") - 1, ingpd.value("txEnd"), self.dir
)
self.fuzzy_junctions[-1].right.get_payload()["end"].set_payload([])
self.fuzzy_junctions[-1].right.get_payload()["end"].get_payload().append(ingpd.value("txEnd"))
# add fuzzy starts
self.start = Bed(ingpd.value("chrom"), ingpd.value("txStart"), ingpd.value("txStart") + 1, self.dir)
self.start.set_payload([])
self.start.get_payload().append(ingpd.value("txStart") + 1)
self.end = Bed(ingpd.value("chrom"), ingpd.value("txEnd") - 1, ingpd.value("txEnd"), self.dir)
self.end.set_payload([])
self.end.get_payload().append(ingpd.value("txEnd"))
def get_loci(transcripts_genepred):
loci = Loci()
loci.verbose= True
with open(transcripts_genepred) as inf:
for line in inf:
if line[0]=='#': continue
gpd = GenePredEntry(line.rstrip())
rng = Bed(gpd.value('chrom'),gpd.value('txStart'),gpd.value('txEnd'))
rng.set_payload(gpd.value('name'))
loc1 = Locus()
loc1.add_member(rng)
loci.add_locus(loc1)
sys.stderr.write("Organizing genepred data into overlapping loci\n")
sys.stderr.write("Started with "+str(len(loci.loci))+" loci\n")
loci.update_loci()
sys.stderr.write("Ended with "+str(len(loci.loci))+" loci\n")
m = 0
locus2name = {}
name2locus = {}
for locus in loci.loci:
m+=1
for member in locus.members:
name = member.get_payload()
if m not in locus2name: locus2name[m] = set()
locus2name[m].add(name)
name2locus[name] = m
return [locus2name,name2locus]
def get_mode(self):
m1 = mode(self.left.get_payload()['junc'])
m2 = mode(self.right.get_payload()['junc'])
return [
Bed(self.chr, m1 - 1, m1, self.dir),
Bed(self.chr, m2 - 1, m2, self.dir)
]
def copy(self):
g = FuzzyGenePred() # start with a blank one why not
# get the settings
for pname in self.params:
g.params[pname] = self.params[pname]
# copy the genepreds
for orig in self.gpds:
g.gpds.append(GenePredEntry(orig.get_line()))
#store direction
g.dir = self.dir
# copy the fuzzy junctions
for orig in self.fuzzy_junctions:
g.fuzzy_junctions.append(orig.copy())
# copy the simple junction set
for orig in self.simple_junction_set:
g.simple_junction_set.add(orig)
# copy the start
if self.start:
g.start = Bed(self.start.chr,\
self.start.start-1,\
self.start.end,\
self.start.direction)
g.start.set_payload([])
for v in self.start.get_payload():
g.start.get_payload().append(v)
# copy the end
if self.end:
g.end = Bed(self.end.chr, self.end.start - 1, self.end.end,
self.end.direction)
g.end.set_payload([])
for v in self.end.get_payload():
g.end.get_payload().append(v)
return g
def do_add_single_exon_fuzzy_gpd(self, fuz2):
if not self.params["do_add_single_exon"]:
return False # make sure we are allowed to be doing this
# build the bounds from the average start and end
s1 = mean(self.start.get_payload())
e1 = mean(self.end.get_payload())
s2 = mean(fuz2.start.get_payload())
e2 = mean(fuz2.end.get_payload())
l1 = e1 - s1 + 1
l2 = e2 - s2 + 1
if l1 < self.params["single_exon_minimum_length"]:
return False
if l2 < self.params["single_exon_minimum_length"]:
return False
if l1 < 1 or l2 < 1:
return False # shouldn't happen
chr1 = self.start.chr
chr2 = self.end.chr
if chr1 != chr2:
return False # shouldn't happen
r1 = Bed(chr1, s1 - 1, e1, self.dir)
r2 = Bed(chr2, s2 - 1, e2, self.dir)
over = r1.overlap_size(r2)
if over < self.params["single_exon_minimum_overlap_bases"]:
return False
# print r1.get_range_string()
# print r2.get_range_string()
cov = min(float(over) / float(l1), float(over) / float(l2))
if cov < self.params["single_exon_minimum_overlap_fraction"]:
return False
if abs(e1 - e2) > self.params["single_exon_maximum_endpoint_distance"]:
return False
if abs(s1 - s2) > self.params["single_exon_maximum_endpoint_distance"]:
return False
# If we're still here, we can add result
output = self.copy()
newstart = output.start.merge(fuz2.start)
newstart.set_payload([])
for s in output.start.get_payload():
newstart.get_payload().append(s)
for s in fuz2.start.get_payload():
newstart.get_payload().append(s)
newend = output.end.merge(fuz2.end)
newend.set_payload([])
for e in output.end.get_payload():
newend.get_payload().append(e)
for e in fuz2.end.get_payload():
newend.get_payload().append(e)
output.start = newstart
output.end = newend
for gpd in fuz2.gpds:
output.gpds.append(gpd)
sjun = get_simple_junction(gpd)
if sjun:
output.simple_junction_set.add(gpd)
return output
def do_add_single_exon_fuzzy_gpd(self, fuz2):
if not self.params['do_add_single_exon']:
return False # make sure we are allowed to be doing this
#build the bounds from the average start and end
s1 = mean(self.start.get_payload())
e1 = mean(self.end.get_payload())
s2 = mean(fuz2.start.get_payload())
e2 = mean(fuz2.end.get_payload())
l1 = e1 - s1 + 1
l2 = e2 - s2 + 1
if l1 < self.params['single_exon_minimum_length']:
return False
if l2 < self.params['single_exon_minimum_length']:
return False
if l1 < 1 or l2 < 1: return False #shouldn't happen
chr1 = self.start.chr
chr2 = self.end.chr
if chr1 != chr2: return False #shouldn't happen
r1 = Bed(chr1, s1 - 1, e1, self.dir)
r2 = Bed(chr2, s2 - 1, e2, self.dir)
over = r1.overlap_size(r2)
if over < self.params['single_exon_minimum_overlap_bases']:
return False
#print r1.get_range_string()
#print r2.get_range_string()
cov = min(float(over) / float(l1), float(over) / float(l2))
if cov < self.params['single_exon_minimum_overlap_fraction']:
return False
if abs(e1 - e2) > self.params['single_exon_maximum_endpoint_distance']:
return False
if abs(s1 - s2) > self.params['single_exon_maximum_endpoint_distance']:
return False
#If we're still here, we can add result
output = self.copy()
newstart = output.start.merge(fuz2.start)
newstart.set_payload([])
for s in output.start.get_payload():
newstart.get_payload().append(s)
for s in fuz2.start.get_payload():
newstart.get_payload().append(s)
newend = output.end.merge(fuz2.end)
newend.set_payload([])
for e in output.end.get_payload():
newend.get_payload().append(e)
for e in fuz2.end.get_payload():
newend.get_payload().append(e)
output.start = newstart
output.end = newend
for gpd in fuz2.gpds:
output.gpds.append(gpd)
sjun = get_simple_junction(gpd)
if sjun:
output.simple_junction_set.add(gpd)
return output
def get_info_string(self):
ostr = ""
ostr += "== FUZZY GENEPRED INFO ==" + "\n"
ostr += str(len(self.gpds)) + " total GPDs" + "\n"
totalbounds = Bed(self.start.chr, self.start.start - 1, self.end.end, self.start.direction)
ostr += totalbounds.get_range_string() + " total bounds\n"
ostr += "---- start ----" + "\n"
ostr += str(len(self.start.get_payload())) + " reads supporting start" + "\n"
ostr += " " + str(mean(self.start.get_payload())) + " mean" + "\n"
ostr += " " + str(mode(self.start.get_payload())) + " mode" + "\n"
ostr += " " + self.start.get_range_string() + " start range\n"
ostr += "---- end ----" + "\n"
ostr += str(len(self.end.get_payload())) + " reads supporting end" + "\n"
ostr += " " + str(mean(self.end.get_payload())) + " mean" + "\n"
ostr += " " + str(mode(self.end.get_payload())) + " mode" + "\n"
ostr += " " + self.end.get_range_string() + " end range\n"
ostr += "---- junctions ----" + "\n"
ostr += str(len(self.fuzzy_junctions)) + " total fuzzy junctions" + "\n"
cnt = 0
for j in self.fuzzy_junctions:
cnt += 1
ostr += (
" "
+ str(cnt)
+ ". "
+ str(mode(j.left.get_payload()["junc"]))
+ " ^ "
+ str(mode(j.right.get_payload()["junc"]))
+ "\n"
)
ostr += " " + j.left.get_range_string() + " ^ " + j.right.get_range_string() + "\n"
ostr += " " + str(len(j.left.get_payload()["junc"])) + " read support" + "\n"
if j.left.get_payload()["start"]:
ostr += " " + "---starts----" + "\n"
ostr += (
" "
+ str(len(j.left.get_payload()["start"].get_payload()))
+ " starts at "
+ j.left.get_payload()["start"].get_range_string()
+ "\n"
)
if j.right.get_payload()["end"]:
ostr += " " + "---ends----" + "\n"
ostr += (
" "
+ str(len(j.right.get_payload()["end"].get_payload()))
+ " ends at "
+ j.right.get_payload()["end"].get_range_string()
+ "\n"
)
return ostr
def target_distance(self, psl_entry, use_direction=False):
if self.value('tName') != psl_entry.value('tName'):
return -1
if use_direction and self.value('strand') != psl_entry.value('strand'):
return -1
b1 = Bed(self.entry['tName'], self.entry['tStart'], self.entry['tEnd'])
b2 = Bed(psl_entry.entry['tName'], psl_entry.entry['tStart'],
psl_entry.entry['tEnd'])
if b1.overlaps(b2):
return 0
if b1.end < b2.start:
return b2.start - b1.end - 1
if b1.start > b2.end:
return b1.start - b2.end - 1
sys.stderr.write("ERROR un accounted for state\n")
sys.exit()
def set_conversion_string(self, conversion_string):
self.conversion_string = conversion_string
self.ars_name = encode_ars_name(conversion_string, self.name)
self.bounds = []
for part in conversion_string.split('/'):
m = re.match('^([^,]+),(\d+)-(\d+)\|([+-])$', part)
self.bounds.append(
Bed(m.group(1), int(m.group(2)), int(m.group(3)), m.group(4)))
def get_info_string(self):
ostr = ''
ostr += "== FUZZY GENEPRED INFO ==" + "\n"
ostr += str(len(self.gpds)) + ' total GPDs' + "\n"
totalbounds = Bed(self.start.chr, self.start.start - 1, self.end.end,
self.start.direction)
ostr += totalbounds.get_range_string() + " total bounds\n"
ostr += '---- start ----' + "\n"
ostr += str(len(
self.start.get_payload())) + " reads supporting start" + "\n"
ostr += ' ' + str(mean(self.start.get_payload())) + ' mean' + "\n"
ostr += ' ' + str(mode(self.start.get_payload())) + ' mode' + "\n"
ostr += ' ' + self.start.get_range_string() + " start range\n"
ostr += '---- end ----' + "\n"
ostr += str(len(
self.end.get_payload())) + " reads supporting end" + "\n"
ostr += ' ' + str(mean(self.end.get_payload())) + ' mean' + "\n"
ostr += ' ' + str(mode(self.end.get_payload())) + ' mode' + "\n"
ostr += ' ' + self.end.get_range_string() + " end range\n"
ostr += '---- junctions ----' + "\n"
ostr += str(len(
self.fuzzy_junctions)) + ' total fuzzy junctions' + "\n"
cnt = 0
for j in self.fuzzy_junctions:
cnt += 1
ostr += ' ' + str(cnt) + '. ' + str(
mode(j.left.get_payload()['junc'])) + " ^ " + str(
mode(j.right.get_payload()['junc'])) + "\n"
ostr += " " + j.left.get_range_string(
) + " ^ " + j.right.get_range_string() + "\n"
ostr += " " + str(len(
j.left.get_payload()['junc'])) + " read support" + "\n"
if j.left.get_payload()['start']:
ostr += " " + "---starts----" + "\n"
ostr += " " + str(
len(j.left.get_payload()['start'].get_payload())
) + " starts at " + j.left.get_payload(
)['start'].get_range_string() + "\n"
if j.right.get_payload()['end']:
ostr += " " + "---ends----" + "\n"
ostr += " " + str(
len(j.right.get_payload()['end'].get_payload())
) + " ends at " + j.right.get_payload(
)['end'].get_range_string() + "\n"
return ostr
def get_beds_from_entry(entry, use_direction=False):
query_beds = []
target_beds = []
print entry
for i in range(0, entry['blockCount']):
if use_direction:
tb = Bed(entry['tName'], entry['tStarts'][i],
entry['tStarts'][i] + entry['blockSizes'][i],
entry['strand'])
target_beds.append(tb)
else:
tb = Bed(entry['tName'], entry['tStarts'][i],
entry['tStarts'][i] + entry['blockSizes'][i])
target_beds.append(tb)
qb = Bed(entry['qName'], entry['qStarts_actual'][i],
entry['qStarts_actual'][i] + entry['blockSizes'][i])
query_beds.append(tb)
return [query_beds, target_beds]
def window_break(inranges, window_size):
outputs = []
if len(inranges) == 0: return outputs
for inrange in inranges:
start = inrange.start
while start + window_size < inrange.end:
outputs.append(Bed(inrange.chr, start, start + window_size - 1))
start += window_size
return outputs
def get_query_bed(self):
s1 = self.value('qStarts_actual')[0]
s2 = self.value('qStarts_actual')[-1] + self.value('blockSizes')[-1]
if self.value('strand') == '-':
s1 = self.convert_coordinate_query_to_actual_query(
self.value('qStarts')[-1] + self.value('blockSizes')[-1]) - 1
s2 = self.convert_coordinate_query_to_actual_query(
self.value('qStarts')[0] + 1)
return Bed(self.value('qName'), s1, s2)
def add_junction(self, inchr, inleft, inright, indir=None):
if not self.left: # this is our first one
t1 = {}
t1["junc"] = []
t1["start"] = None
self.left = Bed(inchr, inleft - 1, inleft, indir)
self.left.set_payload(t1)
self.left.get_payload()["junc"].append(inleft)
self.right = Bed(inchr, inright - 1, inright, indir)
t2 = {}
t2["junc"] = []
t2["end"] = None
self.right = Bed(inchr, inright - 1, inright, indir)
self.right.set_payload(t2)
self.right.get_payload()["junc"].append(inright)
return
# Lets add this one to our current one
newfuz = FuzzyJunction(inchar, inleft, inright, indir)
self.add_fuzzy_junction(newfuz)
def query_overlap_size(self, psl2):
if self.value('qName') != psl2.value('qName'):
return 0
# on same query
output = 0
for i in range(0, self.value('blockCount')):
for j in range(0, psl2.value('blockCount')):
b1 = Bed(
self.value('qName'),
self.value('qStarts_actual')[i],
self.value('qStarts_actual')[i] +
self.value('blockSizes')[i])
b2 = Bed(
psl2.value('qName'),
psl2.value('qStarts_actual')[j],
psl2.value('qStarts_actual')[j] +
psl2.value('blockSizes')[j])
size = b1.overlap_size(b2)
output += size
return output
def target_overlap_size(self, psl2, use_direction=False):
if self.value('tName') != psl2.value('tName'):
return 0
if use_direction and self.value('strand') != psl2.value('strand'):
return 0
# on same chromosome
output = 0
for i in range(0, self.value('blockCount')):
for j in range(0, psl2.value('blockCount')):
b1 = Bed(
self.value('tName'),
self.value('tStarts')[i],
self.value('tStarts')[i] + self.value('blockSizes')[i])
b2 = Bed(
psl2.value('tName'),
psl2.value('tStarts')[j],
psl2.value('tStarts')[j] + psl2.value('blockSizes')[j])
size = b1.overlap_size(b2)
output += size
return output
def copy(self):
newjunc = FuzzyJunction()
newjunc.chr = self.chr
newjunc.left = Bed(self.left.chr,\
self.left.start-1,\
self.left.end,\
self.left.direction)
t1 = {}
t1['junc'] = []
t1['start'] = None
newjunc.left.set_payload(t1)
for j in self.left.get_payload()['junc']:
newjunc.left.get_payload()['junc'].append(j)
newjunc.right = Bed(self.right.chr, self.right.start - 1,
self.right.end, self.right.direction)
#copy any starts for the junction
if self.left.get_payload()['start']:
ls = self.left.get_payload()['start']
newjunc.left.get_payload()['start'] = Bed(ls.chr, ls.start - 1,
ls.end, ls.direction)
newjunc.left.get_payload()['start'].set_payload([])
for p in self.left.get_payload()['start'].get_payload():
newjunc.left.get_payload()['start'].get_payload().append(p)
t2 = {}
t2['junc'] = []
t2['end'] = None
newjunc.right.set_payload(t2)
for j in self.right.get_payload()['junc']:
newjunc.right.get_payload()['junc'].append(j)
#copy any ends for the junction
if self.right.get_payload()['end']:
ren = self.right.get_payload()['end']
newjunc.right.get_payload()['end'] = Bed(ren.chr, ren.start - 1,
ren.end, ren.direction)
newjunc.right.get_payload()['end'].set_payload([])
for p in self.right.get_payload()['end'].get_payload():
newjunc.right.get_payload()['end'].get_payload().append(p)
return newjunc
def main():
parser = argparse.ArgumentParser(description='Create artifical reference sequences from a genepred')
parser.add_argument('gpd_file')
parser.add_argument('reference_fasta')
parser.add_argument('-o','--output',help="output file to write to or STDOUT if not set")
args = parser.parse_args()
of = sys.stdout
if args.output: of = open(args.output,'w')
f = read_fasta_into_hash(args.reference_fasta)
with open(args.gpd_file) as inf:
for line in inf:
gpd = GenePredBasics.GenePredEntry()
gpd.line_to_entry(line.rstrip())
ars = ARS()
beds = []
for i in range(0,gpd.value('exonCount')):
b = Bed(gpd.value('chrom'),gpd.value('exonStarts')[i],gpd.value('exonEnds')[i],gpd.value('strand'))
beds.append(b)
ars.set_bounds(beds)
ars.set_name(gpd.value('name'))
ars.set_sequence_from_original_reference_hash(f)
of.write(ars.get_fasta())
def add_junction(self, inchr, inleft, inright, indir=None):
if not self.left: # this is our first one
t1 = {}
t1['junc'] = []
t1['start'] = None
self.left = Bed(inchr, inleft - 1, inleft, indir)
self.left.set_payload(t1)
self.left.get_payload()['junc'].append(inleft)
self.right = Bed(inchr, inright - 1, inright, indir)
t2 = {}
t2['junc'] = []
t2['end'] = None
self.right = Bed(inchr, inright - 1, inright, indir)
self.right.set_payload(t2)
self.right.get_payload()['junc'].append(inright)
return
#Lets add this one to our current one
newfuz = FuzzyJunction(inchar, inleft, inright, indir)
self.add_fuzzy_junction(newfuz)
def read_first(self, ingpd):
self.gpds.append(ingpd)
sjun = get_simple_junction(ingpd)
if sjun:
self.simple_junction_set.add(sjun)
if self.params['use_dir']: self.dir = ingpd.value('strand')
# add fuzzy junctions
chr = ingpd.value('chrom')
for i in range(0, len(ingpd.value('exonStarts')) - 1):
self.fuzzy_junctions.append(
FuzzyJunction(chr,
ingpd.value('exonEnds')[i],
ingpd.value('exonStarts')[i + 1] + 1, self.dir))
if len(ingpd.value('exonStarts')) > 1: # we have junctions
self.fuzzy_junctions[0].left.get_payload()['start'] = Bed(
chr, ingpd.value('txStart'),
ingpd.value('txStart') + 1, self.dir)
self.fuzzy_junctions[0].left.get_payload()['start'].set_payload([])
self.fuzzy_junctions[0].left.get_payload()['start'].get_payload(
).append(ingpd.value('txStart') + 1)
self.fuzzy_junctions[-1].right.get_payload()['end'] = Bed(
chr,
ingpd.value('txEnd') - 1, ingpd.value('txEnd'), self.dir)
self.fuzzy_junctions[-1].right.get_payload()['end'].set_payload([])
self.fuzzy_junctions[-1].right.get_payload()['end'].get_payload(
).append(ingpd.value('txEnd'))
# add fuzzy starts
self.start = Bed(ingpd.value('chrom'), ingpd.value('txStart'),
ingpd.value('txStart') + 1, self.dir)
self.start.set_payload([])
self.start.get_payload().append(ingpd.value('txStart') + 1)
self.end = Bed(ingpd.value('chrom'),
ingpd.value('txEnd') - 1, ingpd.value('txEnd'),
self.dir)
self.end.set_payload([])
self.end.get_payload().append(ingpd.value('txEnd'))
def get_range(self):
endpos = self.value('pos') - 1
for c in self.value('cigar_array'):
if re.match('[MDNX=]', c['op']): endpos += c['val']
return Bed(self.value('rname'),
self.value('pos') - 1, endpos, self.strand())
def main():
parser = argparse.ArgumentParser()
parser.add_argument('gpd_input')
parser.add_argument('bam_input')
parser.add_argument('--intergenic_buffer', default=10000, type=int)
parser.add_argument('--window_size', default=10000, type=int)
parser.add_argument('--bin_size', default=1000, type=int)
parser.add_argument(
'--use_off_regions',
action='store_true',
help="Use a region even if there is no reads mapped to it.")
parser.add_argument('--get_exons', action='store_true')
args = parser.parse_args()
chr_beds = {}
gene_beds = []
exon_beds = []
sys.stderr.write("Reading genepred file\n")
asum = 0
atot = 0
with open(args.gpd_input) as inf:
for line in inf:
g = GenePredEntry(line)
asum += g.length()
atot += 1
grng = g.get_bed()
grng.direction = None
if grng.chr not in chr_beds:
chr_beds[grng.chr] = grng.copy()
chr_beds[grng.chr] = chr_beds[grng.chr].merge(grng)
gene_beds.append(grng)
for i in range(0, g.get_exon_count()):
erng = Bed(g.value('chrom'),
g.value('exonStarts')[i],
g.value('exonEnds')[i])
exon_beds.append(erng)
avglen = float(asum) / float(atot)
sys.stderr.write("Sorting gene bed\n")
gene_beds = sort_ranges(gene_beds)
gene_beds = merge_ranges(gene_beds, already_sorted=True)
sys.stderr.write("Sorting chromosome beds\n")
chr_beds = sort_ranges([chr_beds[x] for x in chr_beds.keys()])
sys.stderr.write("Sorting exon beds\n")
exon_beds = sort_ranges(exon_beds)
sys.stderr.write("Get padded genes\n")
padded_gene_beds = pad_ranges(gene_beds, args.intergenic_buffer, chr_beds)
padded_gene_beds = merge_ranges(padded_gene_beds, already_sorted=True)
sys.stderr.write("Get intergenic regions\n")
intergenic_beds = subtract_ranges(chr_beds,
padded_gene_beds,
already_sorted=True)
intergenic_beds = merge_ranges(intergenic_beds, already_sorted=True)
intergenic_beds = window_break(intergenic_beds, args.window_size)
#for i in intergenic_beds: print i.get_range_string()
sys.stderr.write("Get merged exons\n")
exon_beds = merge_ranges(exon_beds)
sys.stderr.write("Get introns\n")
intron_beds = subtract_ranges(gene_beds, exon_beds, already_sorted=True)
intron_beds = merge_ranges(intron_beds, already_sorted=True)
intron_beds = window_break(intron_beds, args.window_size)
sys.stderr.write("Going through short reads\n")
cmd = "sam_to_bed_depth.py " + args.bam_input
p = Popen(cmd.split(), stdout=PIPE)
for x in intron_beds:
x.set_payload([]) # payloads are read depths
for x in intergenic_beds:
x.set_payload([]) # payloads are read depths
for x in exon_beds:
x.set_payload([]) # payloads are read depths
introndepth = []
intergenicdepth = []
exondepth = []
pseudoreadcount = 0
if not args.get_exons: exon_beds = []
section_count = 0
while True:
section_count += 1
line = p.stdout.readline()
if not line: break
f = line.split("\t")
depth = int(f[3])
curr = Bed(f[0], int(f[1]), int(f[2]))
if section_count % 100 == 0:
sys.stderr.write(curr.get_range_string() + " \r")
pseudoreadcount += depth
if len(exon_beds) > 0:
while curr.cmp(exon_beds[0]) > 0 and len(
exon_beds) > 0: # we've passed the region
v = exon_beds.pop(0)
if len(v.get_payload()) == 0 and not args.use_off_regions:
continue
av = average(v)
exondepth.append(av)
#print str(av)+" exonic "+v.get_range_string()
c = curr.cmp(exon_beds[0])
if c == 0: # overlaps with intron
size = curr.overlap_size(exon_beds[0])
for i in range(0, size):
exon_beds[0].get_payload().append(depth)
if len(intron_beds) > 0:
while curr.cmp(intron_beds[0]) > 0 and len(
intron_beds) > 0: # we've passed the region
v = intron_beds.pop(0)
if len(v.get_payload()) == 0 and not args.use_off_regions:
continue
av = average(v)
introndepth.append(av)
#print str(av)+" intronic "+v.get_range_string()
c = curr.cmp(intron_beds[0])
if c == 0: # overlaps with intron
size = curr.overlap_size(intron_beds[0])
for i in range(0, size):
intron_beds[0].get_payload().append(depth)
if len(intergenic_beds) > 0:
while curr.cmp(intergenic_beds[0]) > 0 and len(
intergenic_beds) > 0: # we've passed the region
v = intergenic_beds.pop(0)
if len(v.get_payload()) == 0 and not args.use_off_regions:
continue
av = average(v)
intergenicdepth.append(av)
display(curr, introndepth, intergenicdepth, pseudoreadcount,
avglen)
#print str(av)+" intergenic "+v.get_range_string()
c = curr.cmp(intergenic_beds[0])
if c == 0: # overlaps with intron
size = curr.overlap_size(intergenic_beds[0])
for i in range(0, size):
intergenic_beds[0].get_payload().append(depth)
#if c > 0: # we passed the intron
# v = intergenic_beds.pop(0)
# av = average(v)
# intergenicdepth.append(av)
# print str(av)+" intergenic "+v.get_range_string()
if args.use_off_regions:
for x in exon_beds:
introndepth.append(average(x.get_payload()))
for x in intron_beds:
introndepth.append(average(x.get_payload()))
for x in intergenic_beds:
intergenicdepth.append(average(x.get_payload()))
p.communicate()
def get_target_bed(self):
return Bed(self.value('tName'), self.value('tStart'),
self.value('tEnd'), self.value('strand'))
class FuzzyGenePred:
#set use_dir true if you want to use direction and make it direction specific
#set proper_set false if you want to do awesome extending that doesn't really work yet
def __init__(self, ingpd=None, params=None, juntol=10):
# Here is the basic data
self.fuzzy_junctions = []
self.gpds = [] #contributing member genepreds
self.start = None
self.end = None
self.dir = None
# Higher level data
self.simple_junction_set = set(
) # quickly search for if a multi exon gene has been added
#Here is the parameters
self.params = {}
self.params['use_dir'] = False
self.params['junction_tolerance'] = juntol
#Not fully implemented. Do we require a full length match
self.params['proper_set'] = True
# Define thresholds for overlapping single exons
self.params['do_add_single_exon'] = True
self.params['single_exon_minimum_length'] = 200
self.params[
'single_exon_minimum_overlap_fraction'] = 0.8 #reciprocal ... must be this fraction or more on both
self.params[
'single_exon_minimum_overlap_bases'] = 1 #minimum number of bases
self.params['single_exon_maximum_endpoint_distance'] = 1000
if params:
for pname in params:
self.params[pname] = params[pname]
if ingpd:
self.add_gpd(ingpd)
def get_genepred_line(self,
end_select='extremes',
junction_select='mode',
name=None):
if not name:
name = 'fuzGPD_' + random_string(8) + '_' + str(
len(self.fuzzy_junctions) + 1) + '_' + str(len(self.gpds))
ostr = ''
ostr += name + "\t"
ostr += name + "\t"
ostr += self.start.chr + "\t"
ostr += self.gpds[0].value('strand') + "\t"
ostr += str(self.start.start - 1) + "\t"
ostr += str(self.end.end) + "\t"
ostr += str(self.start.start - 1) + "\t"
ostr += str(self.end.end) + "\t"
ostr += str(len(self.fuzzy_junctions) + 1) + "\t"
exonstarts = []
exonends = []
exonstarts.append(self.start.start - 1)
for j in self.fuzzy_junctions:
exonends.append(mode(j.left.get_payload()['junc']))
exonstarts.append(mode(j.right.get_payload()['junc']) - 1)
exonends.append(self.end.end)
ostr += ','.join([str(x) for x in exonstarts]) + ',' + "\t"
ostr += ','.join([str(x) for x in exonends]) + ','
return ostr
# Return a copy of the fuzzy geneprep
def copy(self):
g = FuzzyGenePred() # start with a blank one why not
# get the settings
for pname in self.params:
g.params[pname] = self.params[pname]
# copy the genepreds
for orig in self.gpds:
g.gpds.append(GenePredEntry(orig.get_line()))
#store direction
g.dir = self.dir
# copy the fuzzy junctions
for orig in self.fuzzy_junctions:
g.fuzzy_junctions.append(orig.copy())
# copy the simple junction set
for orig in self.simple_junction_set:
g.simple_junction_set.add(orig)
# copy the start
if self.start:
g.start = Bed(self.start.chr,\
self.start.start-1,\
self.start.end,\
self.start.direction)
g.start.set_payload([])
for v in self.start.get_payload():
g.start.get_payload().append(v)
# copy the end
if self.end:
g.end = Bed(self.end.chr, self.end.start - 1, self.end.end,
self.end.direction)
g.end.set_payload([])
for v in self.end.get_payload():
g.end.get_payload().append(v)
return g
def exon_count(self):
return len(self.fuzzy_junctions) + 1
def gpd_count(self):
return len(self.gpds)
def get_bed(self):
return Bed(self.start.chr, self.start.start - 1, self.end.end,
self.start.direction)
#This is an inspection tool for a fuzzy gpd
def get_info_string(self):
ostr = ''
ostr += "== FUZZY GENEPRED INFO ==" + "\n"
ostr += str(len(self.gpds)) + ' total GPDs' + "\n"
totalbounds = Bed(self.start.chr, self.start.start - 1, self.end.end,
self.start.direction)
ostr += totalbounds.get_range_string() + " total bounds\n"
ostr += '---- start ----' + "\n"
ostr += str(len(
self.start.get_payload())) + " reads supporting start" + "\n"
ostr += ' ' + str(mean(self.start.get_payload())) + ' mean' + "\n"
ostr += ' ' + str(mode(self.start.get_payload())) + ' mode' + "\n"
ostr += ' ' + self.start.get_range_string() + " start range\n"
ostr += '---- end ----' + "\n"
ostr += str(len(
self.end.get_payload())) + " reads supporting end" + "\n"
ostr += ' ' + str(mean(self.end.get_payload())) + ' mean' + "\n"
ostr += ' ' + str(mode(self.end.get_payload())) + ' mode' + "\n"
ostr += ' ' + self.end.get_range_string() + " end range\n"
ostr += '---- junctions ----' + "\n"
ostr += str(len(
self.fuzzy_junctions)) + ' total fuzzy junctions' + "\n"
cnt = 0
for j in self.fuzzy_junctions:
cnt += 1
ostr += ' ' + str(cnt) + '. ' + str(
mode(j.left.get_payload()['junc'])) + " ^ " + str(
mode(j.right.get_payload()['junc'])) + "\n"
ostr += " " + j.left.get_range_string(
) + " ^ " + j.right.get_range_string() + "\n"
ostr += " " + str(len(
j.left.get_payload()['junc'])) + " read support" + "\n"
if j.left.get_payload()['start']:
ostr += " " + "---starts----" + "\n"
ostr += " " + str(
len(j.left.get_payload()['start'].get_payload())
) + " starts at " + j.left.get_payload(
)['start'].get_range_string() + "\n"
if j.right.get_payload()['end']:
ostr += " " + "---ends----" + "\n"
ostr += " " + str(
len(j.right.get_payload()['end'].get_payload())
) + " ends at " + j.right.get_payload(
)['end'].get_range_string() + "\n"
return ostr
#Add a new gpd return true if successful
#Return false if it didn't work, return the new combined if it worked
def add_gpd(self, ingpd):
if len(self.gpds) == 0: # first one
self.read_first(ingpd)
return self #return ourself if we are adding our first
# more difficult situation where we must try to combine
# See if it can match first before actually adding stuff to it
#if self.
newfuz = FuzzyGenePred(ingpd, params=self.params)
output = self.add_fuzzy_gpd(newfuz)
return output
# combine together compatible overlapping sets
def concat_fuzzy_gpd(self, fuz2):
if len(fuz2.fuzzy_junctions) == 0 and len(self.fuzzy_junctions) != 0:
return False
if len(fuz2.fuzzy_junctions) != 0 and len(self.fuzzy_junctions) == 0:
return False
# Lets work combine the single exon step and exit
if len(fuz2.fuzzy_junctions) == 0 and len(self.fuzzy_junctions) == 0:
return self.do_add_single_exon_fuzzy_gpd(fuz2)
# For now don't add them if one is single exon
if len(self.fuzzy_junctions) == 0 or len(fuz2.fuzzy_junctions) == 0:
return False
# See if its already a subset
easy_subset = False
for simplejunction in fuz2.simple_junction_set:
if simplejunction in self.simple_junction_set:
easy_subset = True
# If its not already a subset look deeper
#1. First we need perfect junctions for a run of them
if not easy_subset:
if not self.compatible_overlap(fuz2): return False
# still here. we will work on combining these
output = self.copy()
# first lets put add any overlapping junctions
for i in range(0, len(output.fuzzy_junctions)):
for j in range(0, len(fuz2.fuzzy_junctions)):
if output.fuzzy_junctions[i].overlaps(
fuz2.fuzzy_junctions[j],
fuz2.params['junction_tolerance']):
output.fuzzy_junctions[i].add_fuzzy_junction(
fuz2.fuzzy_junctions[j])
if j == 0: # put the start in too
if not output.fuzzy_junctions[i].left.get_payload(
)['start']:
output.fuzzy_junctions[i].left.get_payload(
)['start'] = fuz2.start.copy()
else: # merge
starts = output.fuzzy_junctions[
i].left.get_payload()['start'].get_payload()
for v in fuz2.start.get_payload():
starts.append(v)
nrange = output.fuzzy_junctions[
i].left.get_payload()['start'].merge(
fuz2.start)
nrange.set_payload(starts[:])
output.fuzzy_junctions[i].left.get_payload(
)['start'] = nrange
if j == len(
fuz2.fuzzy_junctions) - 1: # put the end in too
if not output.fuzzy_junctions[i].right.get_payload(
)['end']:
output.fuzzy_junctions[i].right.get_payload(
)['end'] = fuz2.end.copy()
else: # merge
ends = output.fuzzy_junctions[i].right.get_payload(
)['end'].get_payload()
for v in fuz2.end.get_payload():
ends.append(v)
nrange = output.fuzzy_junctions[
i].right.get_payload()['end'].merge(fuz2.end)
nrange.set_payload(ends[:])
output.fuzzy_junctions[i].right.get_payload(
)['end'] = nrange
# see if we should build onto the left
leftnum = -1
leftmost = self.fuzzy_junctions[0]
if fuz2.fuzzy_junctions[0].right.end < leftmost.left.start:
for i in range(0, len(fuz2.fuzzy_junctions)):
if fuz2.fuzzy_junctions[i].overlaps(
leftmost, fuz2.params['junction_tolerance']):
leftnum = i
break
#leftnum is now -1 if no additions to the left zero if it starts on the same
if leftnum > 0:
for i in reversed(range(0, leftnum)):
output.fuzzy_junctions.insert(0,
fuz2.fuzzy_junctions[i].copy())
output.start = fuz2.start.copy()
rightnum = -1 # get the right point ... our first one comes after this
rightmost = self.fuzzy_junctions[-1]
if fuz2.fuzzy_junctions[-1].left.start > rightmost.right.end:
for i in reversed(range(0, len(fuz2.fuzzy_junctions))):
if fuz2.fuzzy_junctions[i].overlaps(
rightmost, fuz2.params['junction_tolerance']):
rightnum = i
break
if rightnum != -1:
rightnum += 1
if rightnum < len(fuz2.fuzzy_junctions):
for i in range(rightnum, len(fuz2.fuzzy_junctions)):
output.fuzzy_junctions.append(
fuz2.fuzzy_junctions[i].copy())
output.end = fuz2.end.copy()
#print leftnum
#print rightnum
#print fuz2.params['junction_tolerance']
#print 'combining'
return output
# add together subsets
def add_fuzzy_gpd(self, fuz2):
# see if we can add this fuzzy gpd to another
# We treat single exon genes seprately so if only one of them is
# single exon we can't compare them
if len(fuz2.fuzzy_junctions) == 0 and len(self.fuzzy_junctions) != 0:
return False
if len(fuz2.fuzzy_junctions) != 0 and len(self.fuzzy_junctions) == 0:
return False
# Lets work combine the single exon step and exit
if len(fuz2.fuzzy_junctions) == 0 and len(self.fuzzy_junctions) == 0:
return self.do_add_single_exon_fuzzy_gpd(fuz2)
# For now don't add them if one is single exon
if len(self.fuzzy_junctions) == 0 or len(fuz2.fuzzy_junctions) == 0:
return False
# See if its already a subset
easy_subset = False
for simplejunction in fuz2.simple_junction_set:
if simplejunction in self.simple_junction_set:
easy_subset = True
# If its not already a subset look deeper
#1. First we need perfect junctions for a run of them
if not easy_subset:
if not self.compatible_overlap(fuz2): return False
# still here. we will work on combining these
output = self.copy()
#switch over to working on the output now
# If we are still here we can add the two of them together
# If they have the same starting junction we can add their starting points together
if output.fuzzy_junctions[0].overlaps(
fuz2.fuzzy_junctions[0], output.params['junction_tolerance']):
#print 'samestart'
newstart = output.start.merge(fuz2.start)
newstart.set_payload(output.start.get_payload())
for s in fuz2.start.get_payload():
newstart.get_payload().append(s)
output.start = newstart
# Check if the other one is new start
elif mode(fuz2.fuzzy_junctions[0].left.get_payload()['junc']) < mode(
output.fuzzy_junctions[0].left.get_payload()['junc']):
#print "2 start"
output.start = fuz2.start
elif mode(fuz2.fuzzy_junctions[0].left.get_payload()['junc']) > mode(
output.fuzzy_junctions[0].left.get_payload()['junc']):
True
# #print "1 start"
# #we're good to go
else:
sys.stderr.write("WARNING: strange start case abort merge\n")
return False
# lets work the ends now
if output.fuzzy_junctions[-1].overlaps(
fuz2.fuzzy_junctions[-1], output.params['junction_tolerance']):
#print 'sameend'
newend = output.end.merge(fuz2.end)
newend.set_payload(output.end.get_payload())
for s in fuz2.end.get_payload():
newend.get_payload().append(s)
output.end = newend
# Check if the other one is new start
elif mode(fuz2.fuzzy_junctions[-1].right.get_payload()['junc']) > mode(
output.fuzzy_junctions[-1].right.get_payload()['junc']):
#print "2 end"
output.end = fuz2.end
elif mode(fuz2.fuzzy_junctions[-1].right.get_payload()['junc']) < mode(
output.fuzzy_junctions[-1].right.get_payload()['junc']):
True
# #print "1 end"
# #we're good to go
else:
sys.stderr.write("WARNING: strange end case abort merge\n")
u1 = mode(output.fuzzy_junctions[-1].left.get_payload()['junc'])
u2 = mode(fuz2.fuzzy_junctions[-1].left.get_payload()['junc'])
v1 = mode(output.fuzzy_junctions[-1].right.get_payload()['junc'])
v2 = mode(fuz2.fuzzy_junctions[-1].right.get_payload()['junc'])
sys.stderr.write(str(u1) + "\t" + str(u2) + "\n")
sys.stderr.write(str(v1) + "\t" + str(v2) + "\n")
return False
# now the starts and ends have been updated in output.
# iterate through the junctions.
# check for a left overhang.
numfuz2left = 0
numoutleft = 0
if not output.fuzzy_junctions[0].overlaps(
fuz2.fuzzy_junctions[0], output.params['junction_tolerance']):
# see if we need to add sequences from fuz2
if mode(fuz2.fuzzy_junctions[0].left.get_payload()['junc']) < mode(
output.fuzzy_junctions[0].left.get_payload()['junc']):
#print 'left over2'
i = 0
while not output.fuzzy_junctions[0].overlaps(
fuz2.fuzzy_junctions[i],
output.params['junction_tolerance']) and i < len(
fuz2.fuzzy_junctions):
i += 1
numfuz2left = i # number to push on from the fuz2 and increment in
#print numfuz2left
elif mode(
fuz2.fuzzy_junctions[0].left.get_payload()['junc']) > mode(
output.fuzzy_junctions[0].left.get_payload()['junc']):
#print 'left over1'
i = 0
while not output.fuzzy_junctions[i].overlaps(
fuz2.fuzzy_junctions[0],
output.params['junction_tolerance']) and i < len(
output.fuzzy_junctions):
i += 1
numoutleft = i # number to increment in from output
#print numoutleft
else:
sys.stderr.write("WARNING: strange case \n")
return False
# next we can check how long we have a run of the same
ind1 = numoutleft
ind2 = numfuz2left
overlap_size = 0
while ind1 < len(output.fuzzy_junctions) and ind2 < len(fuz2.fuzzy_junctions) \
and output.fuzzy_junctions[ind1].overlaps(fuz2.fuzzy_junctions[ind2],output.params['junction_tolerance']):
overlap_size += 1
ind1 += 1
ind2 += 1
#print 'overlap size '+str(overlap_size)
numoutright = len(output.fuzzy_junctions) - overlap_size - numoutleft
numfuz2right = len(fuz2.fuzzy_junctions) - overlap_size - numfuz2left
if min(numoutright, numfuz2right) != 0:
sys.stderr.write("WARNING: expected one of them to be zero\n")
#print self.get_info_string()
#print '====================='
#print fuz2.get_info_string()
#sys.exit()
return False
if min(numoutleft, numfuz2left) != 0:
sys.stderr.write("WARNING: expected one of them to be zero\n")
return False
#print numoutright
#print numfuz2right
#print output.fuzzy_junctions[numoutleft].overlaps(fuz2.fuzzy_junctions[numfuz2left],output.junction_tolerance)
#print 'add'
#Now we have what we need to go through and do some updating
#Lets just make new fuzzy junctions
newjuncs = []
for i in range(0, numfuz2left):
newjuncs.append(fuz2.fuzzy_junctions[i])
for i in range(0, numoutleft):
newjuncs.append(output.fuzzy_junctions[i])
#Now we do both down the center
range1 = range(numoutleft, overlap_size + numoutleft)
range2 = range(numfuz2left, overlap_size + numfuz2left)
for i in range(0, len(range1)):
newjuncs.append(output.fuzzy_junctions[range1[i]])
newjuncs[-1].add_fuzzy_junction(fuz2.fuzzy_junctions[range2[i]])
#print i
#Make the right size
for i in range(overlap_size + numfuz2left,
overlap_size + numfuz2left + numfuz2right):
newjuncs.append(fuz2.fuzzy_junctions[i])
for i in range(overlap_size + numoutleft,
overlap_size + numoutleft + numoutright):
newjuncs.append(output.fuzzy_junctions[i])
output.fuzzy_junctions = newjuncs
#print 'adding gpd '+str(len(fuz2.gpds))+' entries'
for g in fuz2.gpds:
output.gpds.append(g)
sjun = get_simple_junction(g)
if sjun:
output.simple_junction_set.add(sjun)
#print 'new entry'
#print self.get_info_string()
return output
def do_add_single_exon_fuzzy_gpd(self, fuz2):
if not self.params['do_add_single_exon']:
return False # make sure we are allowed to be doing this
#build the bounds from the average start and end
s1 = mean(self.start.get_payload())
e1 = mean(self.end.get_payload())
s2 = mean(fuz2.start.get_payload())
e2 = mean(fuz2.end.get_payload())
l1 = e1 - s1 + 1
l2 = e2 - s2 + 1
if l1 < self.params['single_exon_minimum_length']:
return False
if l2 < self.params['single_exon_minimum_length']:
return False
if l1 < 1 or l2 < 1: return False #shouldn't happen
chr1 = self.start.chr
chr2 = self.end.chr
if chr1 != chr2: return False #shouldn't happen
r1 = Bed(chr1, s1 - 1, e1, self.dir)
r2 = Bed(chr2, s2 - 1, e2, self.dir)
over = r1.overlap_size(r2)
if over < self.params['single_exon_minimum_overlap_bases']:
return False
#print r1.get_range_string()
#print r2.get_range_string()
cov = min(float(over) / float(l1), float(over) / float(l2))
if cov < self.params['single_exon_minimum_overlap_fraction']:
return False
if abs(e1 - e2) > self.params['single_exon_maximum_endpoint_distance']:
return False
if abs(s1 - s2) > self.params['single_exon_maximum_endpoint_distance']:
return False
#If we're still here, we can add result
output = self.copy()
newstart = output.start.merge(fuz2.start)
newstart.set_payload([])
for s in output.start.get_payload():
newstart.get_payload().append(s)
for s in fuz2.start.get_payload():
newstart.get_payload().append(s)
newend = output.end.merge(fuz2.end)
newend.set_payload([])
for e in output.end.get_payload():
newend.get_payload().append(e)
for e in fuz2.end.get_payload():
newend.get_payload().append(e)
output.start = newstart
output.end = newend
for gpd in fuz2.gpds:
output.gpds.append(gpd)
sjun = get_simple_junction(gpd)
if sjun:
output.simple_junction_set.add(gpd)
return output
#Return true if these fuzzy genepreds can be added together
def compatible_overlap(self, fingpd):
f1 = self
f2 = fingpd
#### Forget about trying zero exon cases for now
if len(f1.fuzzy_junctions) == 0 or len(f2.fuzzy_junctions) == 0:
return False
#Find all matches
matches = []
for i in range(0, len(f1.fuzzy_junctions)):
for j in range(0, len(f2.fuzzy_junctions)):
if f1.fuzzy_junctions[i].overlaps(
f2.fuzzy_junctions[j],
self.params['junction_tolerance']):
matches.append([i, j])
# This is our matched junctions in f1 and f2
if len(matches) == 0:
return False # Nothing matched.. certainly no overlap
# This is the number of extra exons it would take in the middle of the run (shifts)
if len(set([x[0] - x[1] for x in matches])) != 1: return False
# Lets make sure all our exons are consecutive
if len(matches) > 1:
consec1 = list(
set([
matches[i + 1][0] - matches[i][0]
for i in range(0,
len(matches) - 1)
]))
consec2 = list(
set([
matches[i + 1][1] - matches[i][1]
for i in range(0,
len(matches) - 1)
]))
if len(consec1) != 1: return False
if len(consec2) != 1: return False
if consec1[0] != 1: return False
if consec2[0] != 1: return False
# one of them should be zero
if not (matches[0][1] == 0 or matches[0][0] == 0):
return False
# and one of our last matches should be the last junction
if not (len(f1.fuzzy_junctions) - 1 == matches[-1][0]
or len(f2.fuzzy_junctions) - 1 == matches[-1][1]):
return False
#### most of the time we will probably be looking for a proper set
#### unless we are extending the long read for isoform prediction
if self.params['proper_set']:
# check those last overhangs
# one of the two needs to have the start and end points in the consecutive matches
if (matches[0][0] == 0 and len(f1.fuzzy_junctions)-1 == matches[-1][0]) or \
(matches[0][1] == 0 and len(f2.fuzzy_junctions)-1 == matches[-1][1]):
return True
return False
return True
def read_first(self, ingpd):
self.gpds.append(ingpd)
sjun = get_simple_junction(ingpd)
if sjun:
self.simple_junction_set.add(sjun)
if self.params['use_dir']: self.dir = ingpd.value('strand')
# add fuzzy junctions
chr = ingpd.value('chrom')
for i in range(0, len(ingpd.value('exonStarts')) - 1):
self.fuzzy_junctions.append(
FuzzyJunction(chr,
ingpd.value('exonEnds')[i],
ingpd.value('exonStarts')[i + 1] + 1, self.dir))
if len(ingpd.value('exonStarts')) > 1: # we have junctions
self.fuzzy_junctions[0].left.get_payload()['start'] = Bed(
chr, ingpd.value('txStart'),
ingpd.value('txStart') + 1, self.dir)
self.fuzzy_junctions[0].left.get_payload()['start'].set_payload([])
self.fuzzy_junctions[0].left.get_payload()['start'].get_payload(
).append(ingpd.value('txStart') + 1)
self.fuzzy_junctions[-1].right.get_payload()['end'] = Bed(
chr,
ingpd.value('txEnd') - 1, ingpd.value('txEnd'), self.dir)
self.fuzzy_junctions[-1].right.get_payload()['end'].set_payload([])
self.fuzzy_junctions[-1].right.get_payload()['end'].get_payload(
).append(ingpd.value('txEnd'))
# add fuzzy starts
self.start = Bed(ingpd.value('chrom'), ingpd.value('txStart'),
ingpd.value('txStart') + 1, self.dir)
self.start.set_payload([])
self.start.get_payload().append(ingpd.value('txStart') + 1)
self.end = Bed(ingpd.value('chrom'),
ingpd.value('txEnd') - 1, ingpd.value('txEnd'),
self.dir)
self.end.set_payload([])
self.end.get_payload().append(ingpd.value('txEnd'))
# Have finished reading in the first case
# Pre: another fuzzy gpd
# Post: True if they are all overlapping junctions
def is_equal_fuzzy(self, fuz2, use_direction=False):
if use_direction:
if self.dir != fuz2.dir: return False
if len(self.fuzzy_junctions) < 0: return False
if len(fuz2.fuzzy_junctions) < 0: return False
if len(self.fuzzy_junctions) != len(fuz2.fuzzy_junctions):
return False
for i in range(0, len(self.fuzzy_junctions)):
if not self.fuzzy_junctions[i].overlaps(
fuz2.fuzzy_junctions[i],
self.params['junction_tolerance']):
return False
return True
def get_random_gpds_from_pair(pair, genes, ref):
#print 'gene 1 ('+pair[0]+'): '
j1s = set()
j1chrom = genes[pair[0]][0].value('chrom')
j1starts = []
j1ends = []
j1strand = genes[pair[0]][0].value('strand')
j2s = set()
j2chrom = genes[pair[1]][0].value('chrom')
j2starts = []
j2ends = []
j2strand = genes[pair[1]][0].value('strand')
for gpd in genes[pair[0]]:
if gpd.value('strand') != j1strand: continue
if gpd.value('chrom') != j1chrom: continue
j1starts.append(gpd.value('exonStarts')[0])
j1ends.append(gpd.value('exonEnds')[-1])
for j in gpd.calculate_junctions():
j1s.add(j)
#print 'gene 2 ('+pair[1]+'): '
for gpd in genes[pair[1]]:
if gpd.value('strand') != j2strand: continue
if gpd.value('chrom') != j2chrom: continue
j2starts.append(gpd.value('exonStarts')[0])
j2ends.append(gpd.value('exonEnds')[-1])
for j in gpd.calculate_junctions():
j2s.add(j)
j1shuf = list(j1s)
shuffle(j1shuf)
j2shuf = list(j2s)
shuffle(j2shuf)
#print j1shuf[0]
#print j2shuf[0]
if j1strand == '+':
m = re.match('[^:]+:(\d+)', j1shuf[0])
left = Bed(j1chrom,
min(j1starts) - 500,
int(m.group(1)) + 500, j1strand)
fsite1 = int(m.group(1))
else:
m = re.match('[^:]+:(\d+),[^:]+:(\d+)', j1shuf[0])
left = Bed(j1chrom, int(m.group(2)) - 500, max(j1ends) + 500, j1strand)
fsite1 = int(m.group(2))
if j2strand == '+':
m = re.match('[^:]+:(\d+),[^:]+:(\d+)', j2shuf[0])
right = Bed(j2chrom,
int(m.group(2)) - 500,
max(j2ends) + 500, j2strand)
fsite2 = int(m.group(2))
else:
m = re.match('[^:]+:(\d+),[^:]+:(\d+)', j2shuf[0])
right = Bed(j2chrom,
min(j2starts) - 500,
int(m.group(1)) + 500, j2strand)
fsite2 = int(m.group(1))
#print left.get_range_string()+' '+left.direction
#print right.get_range_string()+' '+right.direction
[leftcomp, rightcomp] = get_compatible_transcripts(genes[pair[0]], fsite1,
genes[pair[1]], fsite2)
#print fsite1
#print fsite2
acf = ACF()
acf.add_bounds(left)
acf.add_bounds(right)
ln = leftcomp[0].value('gene_name')
rn = rightcomp[0].value('gene_name')
site_string = leftcomp[0].value('chrom') + ":" + str(
fsite1) + leftcomp[0].value('strand') + '/' + rightcomp[0].value(
'chrom') + ":" + str(fsite2) + rightcomp[0].value('strand')
ars = ARS(ref=ref,
conversion_string=acf.get_conversion_string(),
name=ln + "," + rn + "," + site_string)
#print ars.conversion_string
#print ars.name
#print ars.get_ars_name()
gpds = make_new_genepreds(leftcomp, fsite1, rightcomp, fsite2, ars)
return [gpds, ars]
def main():
parser = argparse.ArgumentParser()
parser.add_argument('gpd_input')
parser.add_argument('bam_input')
parser.add_argument('--intergenic_buffer',default=10000,type=int)
parser.add_argument('--window_size',default=10000,type=int)
parser.add_argument('--bin_size',default=1000,type=int)
parser.add_argument('--use_off_regions',action='store_true',help="Use a region even if there is no reads mapped to it.")
parser.add_argument('--get_exons',action='store_true')
args = parser.parse_args()
chr_beds = {}
gene_beds = []
exon_beds = []
sys.stderr.write("Reading genepred file\n")
asum = 0
atot = 0
with open(args.gpd_input) as inf:
for line in inf:
g = GenePredEntry(line)
asum += g.length()
atot += 1
grng = g.get_bed()
grng.direction = None
if grng.chr not in chr_beds:
chr_beds[grng.chr] = grng.copy()
chr_beds[grng.chr] = chr_beds[grng.chr].merge(grng)
gene_beds.append(grng)
for i in range(0,g.get_exon_count()):
erng = Bed(g.value('chrom'),g.value('exonStarts')[i],g.value('exonEnds')[i])
exon_beds.append(erng)
avglen = float(asum)/float(atot)
sys.stderr.write("Sorting gene bed\n")
gene_beds = sort_ranges(gene_beds)
gene_beds = merge_ranges(gene_beds,already_sorted=True)
sys.stderr.write("Sorting chromosome beds\n")
chr_beds = sort_ranges([chr_beds[x] for x in chr_beds.keys()])
sys.stderr.write("Sorting exon beds\n")
exon_beds = sort_ranges(exon_beds)
sys.stderr.write("Get padded genes\n")
padded_gene_beds = pad_ranges(gene_beds,args.intergenic_buffer,chr_beds)
padded_gene_beds = merge_ranges(padded_gene_beds,already_sorted=True)
sys.stderr.write("Get intergenic regions\n")
intergenic_beds = subtract_ranges(chr_beds,padded_gene_beds,already_sorted=True)
intergenic_beds = merge_ranges(intergenic_beds,already_sorted=True)
intergenic_beds = window_break(intergenic_beds,args.window_size)
#for i in intergenic_beds: print i.get_range_string()
sys.stderr.write("Get merged exons\n")
exon_beds = merge_ranges(exon_beds)
sys.stderr.write("Get introns\n")
intron_beds = subtract_ranges(gene_beds,exon_beds,already_sorted=True)
intron_beds = merge_ranges(intron_beds,already_sorted=True)
intron_beds = window_break(intron_beds,args.window_size)
sys.stderr.write("Going through short reads\n")
cmd = "sam_to_bed_depth.py "+args.bam_input
p = Popen(cmd.split(),stdout=PIPE)
for x in intron_beds: x.set_payload([]) # payloads are read depths
for x in intergenic_beds: x.set_payload([]) # payloads are read depths
for x in exon_beds: x.set_payload([]) # payloads are read depths
introndepth = []
intergenicdepth = []
exondepth = []
pseudoreadcount = 0
if not args.get_exons: exon_beds = []
section_count = 0
while True:
section_count += 1
line = p.stdout.readline()
if not line: break
f = line.split("\t")
depth = int(f[3])
curr = Bed(f[0],int(f[1]),int(f[2]))
if section_count %100==0: sys.stderr.write(curr.get_range_string()+" \r")
pseudoreadcount += depth
if len(exon_beds) > 0:
while curr.cmp(exon_beds[0]) > 0 and len(exon_beds) > 0: # we've passed the region
v = exon_beds.pop(0)
if len(v.get_payload()) == 0 and not args.use_off_regions: continue
av = average(v)
exondepth.append(av)
#print str(av)+" exonic "+v.get_range_string()
c = curr.cmp(exon_beds[0])
if c == 0: # overlaps with intron
size = curr.overlap_size(exon_beds[0])
for i in range(0,size): exon_beds[0].get_payload().append(depth)
if len(intron_beds) > 0:
while curr.cmp(intron_beds[0]) > 0 and len(intron_beds) > 0: # we've passed the region
v = intron_beds.pop(0)
if len(v.get_payload()) == 0 and not args.use_off_regions: continue
av = average(v)
introndepth.append(av)
#print str(av)+" intronic "+v.get_range_string()
c = curr.cmp(intron_beds[0])
if c == 0: # overlaps with intron
size = curr.overlap_size(intron_beds[0])
for i in range(0,size): intron_beds[0].get_payload().append(depth)
if len(intergenic_beds) > 0:
while curr.cmp(intergenic_beds[0]) > 0 and len(intergenic_beds) > 0: # we've passed the region
v = intergenic_beds.pop(0)
if len(v.get_payload()) == 0 and not args.use_off_regions: continue
av = average(v)
intergenicdepth.append(av)
display(curr,introndepth,intergenicdepth,pseudoreadcount,avglen)
#print str(av)+" intergenic "+v.get_range_string()
c = curr.cmp(intergenic_beds[0])
if c == 0: # overlaps with intron
size = curr.overlap_size(intergenic_beds[0])
for i in range(0,size): intergenic_beds[0].get_payload().append(depth)
#if c > 0: # we passed the intron
# v = intergenic_beds.pop(0)
# av = average(v)
# intergenicdepth.append(av)
# print str(av)+" intergenic "+v.get_range_string()
if args.use_off_regions:
for x in exon_beds: introndepth.append(average(x.get_payload()))
for x in intron_beds: introndepth.append(average(x.get_payload()))
for x in intergenic_beds: intergenicdepth.append(average(x.get_payload()))
p.communicate()
class FuzzyGenePred:
# set use_dir true if you want to use direction and make it direction specific
# set proper_set false if you want to do awesome extending that doesn't really work yet
def __init__(self, ingpd=None, params=None, juntol=10):
# Here is the basic data
self.fuzzy_junctions = []
self.gpds = [] # contributing member genepreds
self.start = None
self.end = None
self.dir = None
# Higher level data
self.simple_junction_set = set() # quickly search for if a multi exon gene has been added
# Here is the parameters
self.params = {}
self.params["use_dir"] = False
self.params["junction_tolerance"] = juntol
# Not fully implemented. Do we require a full length match
self.params["proper_set"] = True
# Define thresholds for overlapping single exons
self.params["do_add_single_exon"] = True
self.params["single_exon_minimum_length"] = 200
self.params[
"single_exon_minimum_overlap_fraction"
] = 0.8 # reciprocal ... must be this fraction or more on both
self.params["single_exon_minimum_overlap_bases"] = 1 # minimum number of bases
self.params["single_exon_maximum_endpoint_distance"] = 1000
if params:
for pname in params:
self.params[pname] = params[pname]
if ingpd:
self.add_gpd(ingpd)
def get_genepred_line(self, end_select="extremes", junction_select="mode", name=None):
if not name:
name = "fuzGPD_" + random_string(8) + "_" + str(len(self.fuzzy_junctions) + 1) + "_" + str(len(self.gpds))
ostr = ""
ostr += name + "\t"
ostr += name + "\t"
ostr += self.start.chr + "\t"
ostr += self.gpds[0].value("strand") + "\t"
ostr += str(self.start.start - 1) + "\t"
ostr += str(self.end.end) + "\t"
ostr += str(self.start.start - 1) + "\t"
ostr += str(self.end.end) + "\t"
ostr += str(len(self.fuzzy_junctions) + 1) + "\t"
exonstarts = []
exonends = []
exonstarts.append(self.start.start - 1)
for j in self.fuzzy_junctions:
exonends.append(mode(j.left.get_payload()["junc"]))
exonstarts.append(mode(j.right.get_payload()["junc"]) - 1)
exonends.append(self.end.end)
ostr += ",".join([str(x) for x in exonstarts]) + "," + "\t"
ostr += ",".join([str(x) for x in exonends]) + ","
return ostr
# Return a copy of the fuzzy geneprep
def copy(self):
g = FuzzyGenePred() # start with a blank one why not
# get the settings
for pname in self.params:
g.params[pname] = self.params[pname]
# copy the genepreds
for orig in self.gpds:
g.gpds.append(GenePredEntry(orig.get_line()))
# store direction
g.dir = self.dir
# copy the fuzzy junctions
for orig in self.fuzzy_junctions:
g.fuzzy_junctions.append(orig.copy())
# copy the simple junction set
for orig in self.simple_junction_set:
g.simple_junction_set.add(orig)
# copy the start
if self.start:
g.start = Bed(self.start.chr, self.start.start - 1, self.start.end, self.start.direction)
g.start.set_payload([])
for v in self.start.get_payload():
g.start.get_payload().append(v)
# copy the end
if self.end:
g.end = Bed(self.end.chr, self.end.start - 1, self.end.end, self.end.direction)
g.end.set_payload([])
for v in self.end.get_payload():
g.end.get_payload().append(v)
return g
def exon_count(self):
return len(self.fuzzy_junctions) + 1
def gpd_count(self):
return len(self.gpds)
def get_bed(self):
return Bed(self.start.chr, self.start.start - 1, self.end.end, self.start.direction)
# This is an inspection tool for a fuzzy gpd
def get_info_string(self):
ostr = ""
ostr += "== FUZZY GENEPRED INFO ==" + "\n"
ostr += str(len(self.gpds)) + " total GPDs" + "\n"
totalbounds = Bed(self.start.chr, self.start.start - 1, self.end.end, self.start.direction)
ostr += totalbounds.get_range_string() + " total bounds\n"
ostr += "---- start ----" + "\n"
ostr += str(len(self.start.get_payload())) + " reads supporting start" + "\n"
ostr += " " + str(mean(self.start.get_payload())) + " mean" + "\n"
ostr += " " + str(mode(self.start.get_payload())) + " mode" + "\n"
ostr += " " + self.start.get_range_string() + " start range\n"
ostr += "---- end ----" + "\n"
ostr += str(len(self.end.get_payload())) + " reads supporting end" + "\n"
ostr += " " + str(mean(self.end.get_payload())) + " mean" + "\n"
ostr += " " + str(mode(self.end.get_payload())) + " mode" + "\n"
ostr += " " + self.end.get_range_string() + " end range\n"
ostr += "---- junctions ----" + "\n"
ostr += str(len(self.fuzzy_junctions)) + " total fuzzy junctions" + "\n"
cnt = 0
for j in self.fuzzy_junctions:
cnt += 1
ostr += (
" "
+ str(cnt)
+ ". "
+ str(mode(j.left.get_payload()["junc"]))
+ " ^ "
+ str(mode(j.right.get_payload()["junc"]))
+ "\n"
)
ostr += " " + j.left.get_range_string() + " ^ " + j.right.get_range_string() + "\n"
ostr += " " + str(len(j.left.get_payload()["junc"])) + " read support" + "\n"
if j.left.get_payload()["start"]:
ostr += " " + "---starts----" + "\n"
ostr += (
" "
+ str(len(j.left.get_payload()["start"].get_payload()))
+ " starts at "
+ j.left.get_payload()["start"].get_range_string()
+ "\n"
)
if j.right.get_payload()["end"]:
ostr += " " + "---ends----" + "\n"
ostr += (
" "
+ str(len(j.right.get_payload()["end"].get_payload()))
+ " ends at "
+ j.right.get_payload()["end"].get_range_string()
+ "\n"
)
return ostr
# Add a new gpd return true if successful
# Return false if it didn't work, return the new combined if it worked
def add_gpd(self, ingpd):
if len(self.gpds) == 0: # first one
self.read_first(ingpd)
return self # return ourself if we are adding our first
# more difficult situation where we must try to combine
# See if it can match first before actually adding stuff to it
# if self.
newfuz = FuzzyGenePred(ingpd, params=self.params)
output = self.add_fuzzy_gpd(newfuz)
return output
# combine together compatible overlapping sets
def concat_fuzzy_gpd(self, fuz2):
if len(fuz2.fuzzy_junctions) == 0 and len(self.fuzzy_junctions) != 0:
return False
if len(fuz2.fuzzy_junctions) != 0 and len(self.fuzzy_junctions) == 0:
return False
# Lets work combine the single exon step and exit
if len(fuz2.fuzzy_junctions) == 0 and len(self.fuzzy_junctions) == 0:
return self.do_add_single_exon_fuzzy_gpd(fuz2)
# For now don't add them if one is single exon
if len(self.fuzzy_junctions) == 0 or len(fuz2.fuzzy_junctions) == 0:
return False
# See if its already a subset
easy_subset = False
for simplejunction in fuz2.simple_junction_set:
if simplejunction in self.simple_junction_set:
easy_subset = True
# If its not already a subset look deeper
# 1. First we need perfect junctions for a run of them
if not easy_subset:
if not self.compatible_overlap(fuz2):
return False
# still here. we will work on combining these
output = self.copy()
# first lets put add any overlapping junctions
for i in range(0, len(output.fuzzy_junctions)):
for j in range(0, len(fuz2.fuzzy_junctions)):
if output.fuzzy_junctions[i].overlaps(fuz2.fuzzy_junctions[j], fuz2.params["junction_tolerance"]):
output.fuzzy_junctions[i].add_fuzzy_junction(fuz2.fuzzy_junctions[j])
if j == 0: # put the start in too
if not output.fuzzy_junctions[i].left.get_payload()["start"]:
output.fuzzy_junctions[i].left.get_payload()["start"] = fuz2.start.copy()
else: # merge
starts = output.fuzzy_junctions[i].left.get_payload()["start"].get_payload()
for v in fuz2.start.get_payload():
starts.append(v)
nrange = output.fuzzy_junctions[i].left.get_payload()["start"].merge(fuz2.start)
nrange.set_payload(starts[:])
output.fuzzy_junctions[i].left.get_payload()["start"] = nrange
if j == len(fuz2.fuzzy_junctions) - 1: # put the end in too
if not output.fuzzy_junctions[i].right.get_payload()["end"]:
output.fuzzy_junctions[i].right.get_payload()["end"] = fuz2.end.copy()
else: # merge
ends = output.fuzzy_junctions[i].right.get_payload()["end"].get_payload()
for v in fuz2.end.get_payload():
ends.append(v)
nrange = output.fuzzy_junctions[i].right.get_payload()["end"].merge(fuz2.end)
nrange.set_payload(ends[:])
output.fuzzy_junctions[i].right.get_payload()["end"] = nrange
# see if we should build onto the left
leftnum = -1
leftmost = self.fuzzy_junctions[0]
if fuz2.fuzzy_junctions[0].right.end < leftmost.left.start:
for i in range(0, len(fuz2.fuzzy_junctions)):
if fuz2.fuzzy_junctions[i].overlaps(leftmost, fuz2.params["junction_tolerance"]):
leftnum = i
break
# leftnum is now -1 if no additions to the left zero if it starts on the same
if leftnum > 0:
for i in reversed(range(0, leftnum)):
output.fuzzy_junctions.insert(0, fuz2.fuzzy_junctions[i].copy())
output.start = fuz2.start.copy()
rightnum = -1 # get the right point ... our first one comes after this
rightmost = self.fuzzy_junctions[-1]
if fuz2.fuzzy_junctions[-1].left.start > rightmost.right.end:
for i in reversed(range(0, len(fuz2.fuzzy_junctions))):
if fuz2.fuzzy_junctions[i].overlaps(rightmost, fuz2.params["junction_tolerance"]):
rightnum = i
break
if rightnum != -1:
rightnum += 1
if rightnum < len(fuz2.fuzzy_junctions):
for i in range(rightnum, len(fuz2.fuzzy_junctions)):
output.fuzzy_junctions.append(fuz2.fuzzy_junctions[i].copy())
output.end = fuz2.end.copy()
# print leftnum
# print rightnum
# print fuz2.params['junction_tolerance']
# print 'combining'
return output
# add together subsets
def add_fuzzy_gpd(self, fuz2):
# see if we can add this fuzzy gpd to another
# We treat single exon genes seprately so if only one of them is
# single exon we can't compare them
if len(fuz2.fuzzy_junctions) == 0 and len(self.fuzzy_junctions) != 0:
return False
if len(fuz2.fuzzy_junctions) != 0 and len(self.fuzzy_junctions) == 0:
return False
# Lets work combine the single exon step and exit
if len(fuz2.fuzzy_junctions) == 0 and len(self.fuzzy_junctions) == 0:
return self.do_add_single_exon_fuzzy_gpd(fuz2)
# For now don't add them if one is single exon
if len(self.fuzzy_junctions) == 0 or len(fuz2.fuzzy_junctions) == 0:
return False
# See if its already a subset
easy_subset = False
for simplejunction in fuz2.simple_junction_set:
if simplejunction in self.simple_junction_set:
easy_subset = True
# If its not already a subset look deeper
# 1. First we need perfect junctions for a run of them
if not easy_subset:
if not self.compatible_overlap(fuz2):
return False
# still here. we will work on combining these
output = self.copy()
# switch over to working on the output now
# If we are still here we can add the two of them together
# If they have the same starting junction we can add their starting points together
if output.fuzzy_junctions[0].overlaps(fuz2.fuzzy_junctions[0], output.params["junction_tolerance"]):
# print 'samestart'
newstart = output.start.merge(fuz2.start)
newstart.set_payload(output.start.get_payload())
for s in fuz2.start.get_payload():
newstart.get_payload().append(s)
output.start = newstart
# Check if the other one is new start
elif mode(fuz2.fuzzy_junctions[0].left.get_payload()["junc"]) < mode(
output.fuzzy_junctions[0].left.get_payload()["junc"]
):
# print "2 start"
output.start = fuz2.start
elif mode(fuz2.fuzzy_junctions[0].left.get_payload()["junc"]) > mode(
output.fuzzy_junctions[0].left.get_payload()["junc"]
):
True
# #print "1 start"
# #we're good to go
else:
sys.stderr.write("WARNING: strange start case abort merge\n")
return False
# lets work the ends now
if output.fuzzy_junctions[-1].overlaps(fuz2.fuzzy_junctions[-1], output.params["junction_tolerance"]):
# print 'sameend'
newend = output.end.merge(fuz2.end)
newend.set_payload(output.end.get_payload())
for s in fuz2.end.get_payload():
newend.get_payload().append(s)
output.end = newend
# Check if the other one is new start
elif mode(fuz2.fuzzy_junctions[-1].right.get_payload()["junc"]) > mode(
output.fuzzy_junctions[-1].right.get_payload()["junc"]
):
# print "2 end"
output.end = fuz2.end
elif mode(fuz2.fuzzy_junctions[-1].right.get_payload()["junc"]) < mode(
output.fuzzy_junctions[-1].right.get_payload()["junc"]
):
True
# #print "1 end"
# #we're good to go
else:
sys.stderr.write("WARNING: strange end case abort merge\n")
u1 = mode(output.fuzzy_junctions[-1].left.get_payload()["junc"])
u2 = mode(fuz2.fuzzy_junctions[-1].left.get_payload()["junc"])
v1 = mode(output.fuzzy_junctions[-1].right.get_payload()["junc"])
v2 = mode(fuz2.fuzzy_junctions[-1].right.get_payload()["junc"])
sys.stderr.write(str(u1) + "\t" + str(u2) + "\n")
sys.stderr.write(str(v1) + "\t" + str(v2) + "\n")
return False
# now the starts and ends have been updated in output.
# iterate through the junctions.
# check for a left overhang.
numfuz2left = 0
numoutleft = 0
if not output.fuzzy_junctions[0].overlaps(fuz2.fuzzy_junctions[0], output.params["junction_tolerance"]):
# see if we need to add sequences from fuz2
if mode(fuz2.fuzzy_junctions[0].left.get_payload()["junc"]) < mode(
output.fuzzy_junctions[0].left.get_payload()["junc"]
):
# print 'left over2'
i = 0
while not output.fuzzy_junctions[0].overlaps(
fuz2.fuzzy_junctions[i], output.params["junction_tolerance"]
) and i < len(fuz2.fuzzy_junctions):
i += 1
numfuz2left = i # number to push on from the fuz2 and increment in
# print numfuz2left
elif mode(fuz2.fuzzy_junctions[0].left.get_payload()["junc"]) > mode(
output.fuzzy_junctions[0].left.get_payload()["junc"]
):
# print 'left over1'
i = 0
while not output.fuzzy_junctions[i].overlaps(
fuz2.fuzzy_junctions[0], output.params["junction_tolerance"]
) and i < len(output.fuzzy_junctions):
i += 1
numoutleft = i # number to increment in from output
# print numoutleft
else:
sys.stderr.write("WARNING: strange case \n")
return False
# next we can check how long we have a run of the same
ind1 = numoutleft
ind2 = numfuz2left
overlap_size = 0
while (
ind1 < len(output.fuzzy_junctions)
and ind2 < len(fuz2.fuzzy_junctions)
and output.fuzzy_junctions[ind1].overlaps(fuz2.fuzzy_junctions[ind2], output.params["junction_tolerance"])
):
overlap_size += 1
ind1 += 1
ind2 += 1
# print 'overlap size '+str(overlap_size)
numoutright = len(output.fuzzy_junctions) - overlap_size - numoutleft
numfuz2right = len(fuz2.fuzzy_junctions) - overlap_size - numfuz2left
if min(numoutright, numfuz2right) != 0:
sys.stderr.write("WARNING: expected one of them to be zero\n")
# print self.get_info_string()
# print '====================='
# print fuz2.get_info_string()
# sys.exit()
return False
if min(numoutleft, numfuz2left) != 0:
sys.stderr.write("WARNING: expected one of them to be zero\n")
return False
# print numoutright
# print numfuz2right
# print output.fuzzy_junctions[numoutleft].overlaps(fuz2.fuzzy_junctions[numfuz2left],output.junction_tolerance)
# print 'add'
# Now we have what we need to go through and do some updating
# Lets just make new fuzzy junctions
newjuncs = []
for i in range(0, numfuz2left):
newjuncs.append(fuz2.fuzzy_junctions[i])
for i in range(0, numoutleft):
newjuncs.append(output.fuzzy_junctions[i])
# Now we do both down the center
range1 = range(numoutleft, overlap_size + numoutleft)
range2 = range(numfuz2left, overlap_size + numfuz2left)
for i in range(0, len(range1)):
newjuncs.append(output.fuzzy_junctions[range1[i]])
newjuncs[-1].add_fuzzy_junction(fuz2.fuzzy_junctions[range2[i]])
# print i
# Make the right size
for i in range(overlap_size + numfuz2left, overlap_size + numfuz2left + numfuz2right):
newjuncs.append(fuz2.fuzzy_junctions[i])
for i in range(overlap_size + numoutleft, overlap_size + numoutleft + numoutright):
newjuncs.append(output.fuzzy_junctions[i])
output.fuzzy_junctions = newjuncs
# print 'adding gpd '+str(len(fuz2.gpds))+' entries'
for g in fuz2.gpds:
output.gpds.append(g)
sjun = get_simple_junction(g)
if sjun:
output.simple_junction_set.add(sjun)
# print 'new entry'
# print self.get_info_string()
return output
def do_add_single_exon_fuzzy_gpd(self, fuz2):
if not self.params["do_add_single_exon"]:
return False # make sure we are allowed to be doing this
# build the bounds from the average start and end
s1 = mean(self.start.get_payload())
e1 = mean(self.end.get_payload())
s2 = mean(fuz2.start.get_payload())
e2 = mean(fuz2.end.get_payload())
l1 = e1 - s1 + 1
l2 = e2 - s2 + 1
if l1 < self.params["single_exon_minimum_length"]:
return False
if l2 < self.params["single_exon_minimum_length"]:
return False
if l1 < 1 or l2 < 1:
return False # shouldn't happen
chr1 = self.start.chr
chr2 = self.end.chr
if chr1 != chr2:
return False # shouldn't happen
r1 = Bed(chr1, s1 - 1, e1, self.dir)
r2 = Bed(chr2, s2 - 1, e2, self.dir)
over = r1.overlap_size(r2)
if over < self.params["single_exon_minimum_overlap_bases"]:
return False
# print r1.get_range_string()
# print r2.get_range_string()
cov = min(float(over) / float(l1), float(over) / float(l2))
if cov < self.params["single_exon_minimum_overlap_fraction"]:
return False
if abs(e1 - e2) > self.params["single_exon_maximum_endpoint_distance"]:
return False
if abs(s1 - s2) > self.params["single_exon_maximum_endpoint_distance"]:
return False
# If we're still here, we can add result
output = self.copy()
newstart = output.start.merge(fuz2.start)
newstart.set_payload([])
for s in output.start.get_payload():
newstart.get_payload().append(s)
for s in fuz2.start.get_payload():
newstart.get_payload().append(s)
newend = output.end.merge(fuz2.end)
newend.set_payload([])
for e in output.end.get_payload():
newend.get_payload().append(e)
for e in fuz2.end.get_payload():
newend.get_payload().append(e)
output.start = newstart
output.end = newend
for gpd in fuz2.gpds:
output.gpds.append(gpd)
sjun = get_simple_junction(gpd)
if sjun:
output.simple_junction_set.add(gpd)
return output
# Return true if these fuzzy genepreds can be added together
def compatible_overlap(self, fingpd):
f1 = self
f2 = fingpd
#### Forget about trying zero exon cases for now
if len(f1.fuzzy_junctions) == 0 or len(f2.fuzzy_junctions) == 0:
return False
# Find all matches
matches = []
for i in range(0, len(f1.fuzzy_junctions)):
for j in range(0, len(f2.fuzzy_junctions)):
if f1.fuzzy_junctions[i].overlaps(f2.fuzzy_junctions[j], self.params["junction_tolerance"]):
matches.append([i, j])
# This is our matched junctions in f1 and f2
if len(matches) == 0:
return False # Nothing matched.. certainly no overlap
# This is the number of extra exons it would take in the middle of the run (shifts)
if len(set([x[0] - x[1] for x in matches])) != 1:
return False
# Lets make sure all our exons are consecutive
if len(matches) > 1:
consec1 = list(set([matches[i + 1][0] - matches[i][0] for i in range(0, len(matches) - 1)]))
consec2 = list(set([matches[i + 1][1] - matches[i][1] for i in range(0, len(matches) - 1)]))
if len(consec1) != 1:
return False
if len(consec2) != 1:
return False
if consec1[0] != 1:
return False
if consec2[0] != 1:
return False
# one of them should be zero
if not (matches[0][1] == 0 or matches[0][0] == 0):
return False
# and one of our last matches should be the last junction
if not (len(f1.fuzzy_junctions) - 1 == matches[-1][0] or len(f2.fuzzy_junctions) - 1 == matches[-1][1]):
return False
#### most of the time we will probably be looking for a proper set
#### unless we are extending the long read for isoform prediction
if self.params["proper_set"]:
# check those last overhangs
# one of the two needs to have the start and end points in the consecutive matches
if (matches[0][0] == 0 and len(f1.fuzzy_junctions) - 1 == matches[-1][0]) or (
matches[0][1] == 0 and len(f2.fuzzy_junctions) - 1 == matches[-1][1]
):
return True
return False
return True
def read_first(self, ingpd):
self.gpds.append(ingpd)
sjun = get_simple_junction(ingpd)
if sjun:
self.simple_junction_set.add(sjun)
if self.params["use_dir"]:
self.dir = ingpd.value("strand")
# add fuzzy junctions
chr = ingpd.value("chrom")
for i in range(0, len(ingpd.value("exonStarts")) - 1):
self.fuzzy_junctions.append(
FuzzyJunction(chr, ingpd.value("exonEnds")[i], ingpd.value("exonStarts")[i + 1] + 1, self.dir)
)
if len(ingpd.value("exonStarts")) > 1: # we have junctions
self.fuzzy_junctions[0].left.get_payload()["start"] = Bed(
chr, ingpd.value("txStart"), ingpd.value("txStart") + 1, self.dir
)
self.fuzzy_junctions[0].left.get_payload()["start"].set_payload([])
self.fuzzy_junctions[0].left.get_payload()["start"].get_payload().append(ingpd.value("txStart") + 1)
self.fuzzy_junctions[-1].right.get_payload()["end"] = Bed(
chr, ingpd.value("txEnd") - 1, ingpd.value("txEnd"), self.dir
)
self.fuzzy_junctions[-1].right.get_payload()["end"].set_payload([])
self.fuzzy_junctions[-1].right.get_payload()["end"].get_payload().append(ingpd.value("txEnd"))
# add fuzzy starts
self.start = Bed(ingpd.value("chrom"), ingpd.value("txStart"), ingpd.value("txStart") + 1, self.dir)
self.start.set_payload([])
self.start.get_payload().append(ingpd.value("txStart") + 1)
self.end = Bed(ingpd.value("chrom"), ingpd.value("txEnd") - 1, ingpd.value("txEnd"), self.dir)
self.end.set_payload([])
self.end.get_payload().append(ingpd.value("txEnd"))
# Have finished reading in the first case
# Pre: another fuzzy gpd
# Post: True if they are all overlapping junctions
def is_equal_fuzzy(self, fuz2, use_direction=False):
if use_direction:
if self.dir != fuz2.dir:
return False
if len(self.fuzzy_junctions) < 0:
return False
if len(fuz2.fuzzy_junctions) < 0:
return False
if len(self.fuzzy_junctions) != len(fuz2.fuzzy_junctions):
return False
for i in range(0, len(self.fuzzy_junctions)):
if not self.fuzzy_junctions[i].overlaps(fuz2.fuzzy_junctions[i], self.params["junction_tolerance"]):
return False
return True
class FuzzyJunction:
# Pre: inleft is 1-indexed last exonic base on the left
# inright is 1-indexed first exonic base on the right
# direction doesn't need to be used
def __init__(self, inchr=None, inleft=None, inright=None, indir=None):
self.chr = inchr
self.left = None #range with payloads being the actual left and rights
self.right = None
self.dir = indir
if inchr and inleft and inright:
self.add_junction(inchr, inleft, inright, indir)
def copy(self):
newjunc = FuzzyJunction()
newjunc.chr = self.chr
newjunc.left = Bed(self.left.chr,\
self.left.start-1,\
self.left.end,\
self.left.direction)
t1 = {}
t1['junc'] = []
t1['start'] = None
newjunc.left.set_payload(t1)
for j in self.left.get_payload()['junc']:
newjunc.left.get_payload()['junc'].append(j)
newjunc.right = Bed(self.right.chr, self.right.start - 1,
self.right.end, self.right.direction)
#copy any starts for the junction
if self.left.get_payload()['start']:
ls = self.left.get_payload()['start']
newjunc.left.get_payload()['start'] = Bed(ls.chr, ls.start - 1,
ls.end, ls.direction)
newjunc.left.get_payload()['start'].set_payload([])
for p in self.left.get_payload()['start'].get_payload():
newjunc.left.get_payload()['start'].get_payload().append(p)
t2 = {}
t2['junc'] = []
t2['end'] = None
newjunc.right.set_payload(t2)
for j in self.right.get_payload()['junc']:
newjunc.right.get_payload()['junc'].append(j)
#copy any ends for the junction
if self.right.get_payload()['end']:
ren = self.right.get_payload()['end']
newjunc.right.get_payload()['end'] = Bed(ren.chr, ren.start - 1,
ren.end, ren.direction)
newjunc.right.get_payload()['end'].set_payload([])
for p in self.right.get_payload()['end'].get_payload():
newjunc.right.get_payload()['end'].get_payload().append(p)
return newjunc
# return chr, and the left and right mode as an array
def get_mode(self):
m1 = mode(self.left.get_payload()['junc'])
m2 = mode(self.right.get_payload()['junc'])
return [
Bed(self.chr, m1 - 1, m1, self.dir),
Bed(self.chr, m2 - 1, m2, self.dir)
]
# Find the mode of the junction and see if this overlaps
def overlaps(self, fjun2, juntol):
m1 = self.get_mode()
m2 = fjun2.get_mode()
if m1[0].chr != m2[0].chr: return False
if m1[0].direction != m2[0].direction:
return False # usually they are both off
if not m1[0].overlaps_with_padding(m2[0], juntol): return False
if not m1[1].overlaps_with_padding(m2[1], juntol): return False
return True
#Right now assumes these are overlap verified prior to calling
def add_junction(self, inchr, inleft, inright, indir=None):
if not self.left: # this is our first one
t1 = {}
t1['junc'] = []
t1['start'] = None
self.left = Bed(inchr, inleft - 1, inleft, indir)
self.left.set_payload(t1)
self.left.get_payload()['junc'].append(inleft)
self.right = Bed(inchr, inright - 1, inright, indir)
t2 = {}
t2['junc'] = []
t2['end'] = None
self.right = Bed(inchr, inright - 1, inright, indir)
self.right.set_payload(t2)
self.right.get_payload()['junc'].append(inright)
return
#Lets add this one to our current one
newfuz = FuzzyJunction(inchar, inleft, inright, indir)
self.add_fuzzy_junction(newfuz)
def add_fuzzy_junction(self, newfuz):
#print 'add fuzzy'
mergeleft = self.left.merge(newfuz.left)
mergeleft.set_payload(self.left.get_payload())
mergeright = self.right.merge(newfuz.right)
mergeright.set_payload(self.right.get_payload())
for j1 in newfuz.left.get_payload()['junc']:
mergeleft.get_payload()['junc'].append(j1)
for j2 in newfuz.right.get_payload()['junc']:
mergeright.get_payload()['junc'].append(j2)
#fix the starts
if newfuz.left.get_payload(
)['start'] and not self.left.get_payload()['start']:
mergeleft.get_payload()['start'] = newfuz.left.get_payload(
)['start']
elif newfuz.left.get_payload()['start'] and self.left.get_payload(
)['start']:
newrange = self.left.get_payload()['start'].merge(
newfuz.left.get_payload()['start'])
newrange.set_payload([])
for s in self.left.get_payload()['start'].get_payload():
newrange.get_payload().append(s)
for s in newfuz.left.get_payload()['start'].get_payload():
newrange.get_payload().append(s)
mergeleft.get_payload()['start'] = newrange
#print 'update left starts'
#fix the ends
if newfuz.right.get_payload(
)['end'] and not self.right.get_payload()['end']:
mergeright.get_payload()['end'] = newfuz.right.get_payload()['end']
elif newfuz.right.get_payload()['end'] and self.right.get_payload(
)['end']:
newrange = newfuz.right.get_payload()['end'].merge(
self.right.get_payload()['end'])
newrange.set_payload([])
for s in self.right.get_payload()['end'].get_payload():
newrange.get_payload().append(s)
for s in newfuz.right.get_payload()['end'].get_payload():
newrange.get_payload().append(s)
mergeright.get_payload()['end'] = newrange
#print 'update right ends'
# We finished the changes
self.left = mergeleft
self.right = mergeright
class FuzzyJunction:
# Pre: inleft is 1-indexed last exonic base on the left
# inright is 1-indexed first exonic base on the right
# direction doesn't need to be used
def __init__(self, inchr=None, inleft=None, inright=None, indir=None):
self.chr = inchr
self.left = None # range with payloads being the actual left and rights
self.right = None
self.dir = indir
if inchr and inleft and inright:
self.add_junction(inchr, inleft, inright, indir)
def copy(self):
newjunc = FuzzyJunction()
newjunc.chr = self.chr
newjunc.left = Bed(self.left.chr, self.left.start - 1, self.left.end, self.left.direction)
t1 = {}
t1["junc"] = []
t1["start"] = None
newjunc.left.set_payload(t1)
for j in self.left.get_payload()["junc"]:
newjunc.left.get_payload()["junc"].append(j)
newjunc.right = Bed(self.right.chr, self.right.start - 1, self.right.end, self.right.direction)
# copy any starts for the junction
if self.left.get_payload()["start"]:
ls = self.left.get_payload()["start"]
newjunc.left.get_payload()["start"] = Bed(ls.chr, ls.start - 1, ls.end, ls.direction)
newjunc.left.get_payload()["start"].set_payload([])
for p in self.left.get_payload()["start"].get_payload():
newjunc.left.get_payload()["start"].get_payload().append(p)
t2 = {}
t2["junc"] = []
t2["end"] = None
newjunc.right.set_payload(t2)
for j in self.right.get_payload()["junc"]:
newjunc.right.get_payload()["junc"].append(j)
# copy any ends for the junction
if self.right.get_payload()["end"]:
ren = self.right.get_payload()["end"]
newjunc.right.get_payload()["end"] = Bed(ren.chr, ren.start - 1, ren.end, ren.direction)
newjunc.right.get_payload()["end"].set_payload([])
for p in self.right.get_payload()["end"].get_payload():
newjunc.right.get_payload()["end"].get_payload().append(p)
return newjunc
# return chr, and the left and right mode as an array
def get_mode(self):
m1 = mode(self.left.get_payload()["junc"])
m2 = mode(self.right.get_payload()["junc"])
return [Bed(self.chr, m1 - 1, m1, self.dir), Bed(self.chr, m2 - 1, m2, self.dir)]
# Find the mode of the junction and see if this overlaps
def overlaps(self, fjun2, juntol):
m1 = self.get_mode()
m2 = fjun2.get_mode()
if m1[0].chr != m2[0].chr:
return False
if m1[0].direction != m2[0].direction:
return False # usually they are both off
if not m1[0].overlaps_with_padding(m2[0], juntol):
return False
if not m1[1].overlaps_with_padding(m2[1], juntol):
return False
return True
# Right now assumes these are overlap verified prior to calling
def add_junction(self, inchr, inleft, inright, indir=None):
if not self.left: # this is our first one
t1 = {}
t1["junc"] = []
t1["start"] = None
self.left = Bed(inchr, inleft - 1, inleft, indir)
self.left.set_payload(t1)
self.left.get_payload()["junc"].append(inleft)
self.right = Bed(inchr, inright - 1, inright, indir)
t2 = {}
t2["junc"] = []
t2["end"] = None
self.right = Bed(inchr, inright - 1, inright, indir)
self.right.set_payload(t2)
self.right.get_payload()["junc"].append(inright)
return
# Lets add this one to our current one
newfuz = FuzzyJunction(inchar, inleft, inright, indir)
self.add_fuzzy_junction(newfuz)
def add_fuzzy_junction(self, newfuz):
# print 'add fuzzy'
mergeleft = self.left.merge(newfuz.left)
mergeleft.set_payload(self.left.get_payload())
mergeright = self.right.merge(newfuz.right)
mergeright.set_payload(self.right.get_payload())
for j1 in newfuz.left.get_payload()["junc"]:
mergeleft.get_payload()["junc"].append(j1)
for j2 in newfuz.right.get_payload()["junc"]:
mergeright.get_payload()["junc"].append(j2)
# fix the starts
if newfuz.left.get_payload()["start"] and not self.left.get_payload()["start"]:
mergeleft.get_payload()["start"] = newfuz.left.get_payload()["start"]
elif newfuz.left.get_payload()["start"] and self.left.get_payload()["start"]:
newrange = self.left.get_payload()["start"].merge(newfuz.left.get_payload()["start"])
newrange.set_payload([])
for s in self.left.get_payload()["start"].get_payload():
newrange.get_payload().append(s)
for s in newfuz.left.get_payload()["start"].get_payload():
newrange.get_payload().append(s)
mergeleft.get_payload()["start"] = newrange
# print 'update left starts'
# fix the ends
if newfuz.right.get_payload()["end"] and not self.right.get_payload()["end"]:
mergeright.get_payload()["end"] = newfuz.right.get_payload()["end"]
elif newfuz.right.get_payload()["end"] and self.right.get_payload()["end"]:
newrange = newfuz.right.get_payload()["end"].merge(self.right.get_payload()["end"])
newrange.set_payload([])
for s in self.right.get_payload()["end"].get_payload():
newrange.get_payload().append(s)
for s in newfuz.right.get_payload()["end"].get_payload():
newrange.get_payload().append(s)
mergeright.get_payload()["end"] = newrange
# print 'update right ends'
# We finished the changes
self.left = mergeleft
self.right = mergeright
def get_bed(self):
return Bed(self.start.chr, self.start.start - 1, self.end.end,
self.start.direction)