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_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 process_read(mpa, args):
# Filter entries by a minimum alignment coverage
newentries = []
for i in [
i for i in range(0, len(mpa.entries))
if mpa.entries[i].get_coverage() > args.minimum_alignment_coverage
]:
newentries.append(mpa.entries[i])
mpa.entries = newentries
# Find best singles
bestsingle = None
bestsinglescore = -1
for i in range(0, len(mpa.entries)):
totalcov = mpa.entries[i].get_coverage()
weightedcov = float(mpa.entries[i].get_coverage()) * float(
mpa.entries[i].get_quality())
if weightedcov > bestsinglescore:
bestsinglescore = weightedcov
bestsingle = i
if bestsinglescore == -1:
sys.stderr.write("failed to find a single path\n")
return None
my_max_intron = args.maximum_intron
if args.fusion: my_max_intron = -1 # we can look any distance for a group
mpa.compatible_graph(
max_intron=my_max_intron,
max_query_overlap=args.maximum_query_overlap,
max_gap=args.maximum_query_gap,
max_target_overlap=args.maximum_target_overlap,
max_query_fraction_overlap=args.maximum_query_fraction_overlap)
ps = mpa.get_root_paths()
bestpath = [bestsingle]
bestscore = 0
besttotalcov = 0
allscores = []
allcov = []
best_path_index = -1
zz = 0
for path in ps:
totalcov = sum([mpa.entries[i].get_coverage() for i in path])
weightedcov = sum([
float(mpa.entries[i].get_coverage()) *
float(mpa.entries[i].get_quality()) for i in path
])
allscores.append(weightedcov)
allcov.append(totalcov)
if weightedcov > bestscore:
bestscore = weightedcov
bestpath = path
besttotalcov = totalcov
best_path_index = zz
zz += 1
#if not bestpath: return None
otherpaths = []
for i in range(0, len(ps)):
if i != best_path_index:
otherpaths.append(ps[i])
query_target_coverages = []
for other_path in otherpaths:
qcov = 0
tcov = 0
for other_entry in [mpa.entries[i] for i in other_path]:
for entry in [mpa.entries[j] for j in bestpath]:
qcov += other_entry.query_overlap_size(entry)
tcov += other_entry.target_overlap_size(entry)
query_target_coverages.append(str(qcov) + '/' + str(tcov))
gapsizes = []
if len(bestpath) > 1:
gapsizes = [
mpa.entries[bestpath[j + 1]].get_query_bed().start -
mpa.entries[bestpath[j]].get_query_bed().end - 1
for j in range(0,
len(bestpath) - 1)
]
#print mpa.g.get_status_string()
#print [mpa.entries[i].get_target_bed().get_range_string() for i in bestpath]
#print [mpa.entries[i].get_query_bed().get_range_string() for i in bestpath]
#print [mpa.entries[i].get_quality() for i in bestpath]
#print [mpa.entries[i].get_coverage() for i in bestpath]
#print gapsizes
#print bestscore
#print bestsinglescore
#See if we should use the single path score instead
if len(path) > 1 and bestsinglescore * (
1 + args.multipath_score_improvement) > bestscore:
bestpath = [bestsingle]
besttotalcov = mpa.entries[bestsingle].get_coverage()
bestscore = bestsinglescore
query_span = mpa.entries[bestpath[0]].get_query_bed()
loci = Loci()
loci.set_use_direction(True)
loci.set_minimum_distance(args.maximum_intron)
for i in bestpath:
r = mpa.entries[i].get_target_bed()
locus = Locus()
locus.set_use_direction(True)
locus.add_member(r)
loci.add_locus(locus)
loci.update_loci()
if len(bestpath) > 1:
for i in bestpath[1:]:
query_span = mpa.entries[i].get_query_bed().merge(query_span)
report = ''
report += mpa.entries[bestpath[0]].value('qName') + "\t"
report += str(len(bestpath)) + "\t"
report += str(len(loci.loci)) + "\t"
report += query_span.get_range_string() + "\t"
report += ','.join([mpa.entries[i].value('strand')
for i in bestpath]) + "\t"
report += ','.join(
[mpa.entries[i].get_query_bed().get_range_string()
for i in bestpath]) + "\t"
report += ','.join(
[mpa.entries[i].get_target_bed().get_range_string()
for i in bestpath]) + "\t"
report += ','.join([str(mpa.entries[i].get_quality())
for i in bestpath]) + "\t"
report += ','.join([str(mpa.entries[i].get_coverage())
for i in bestpath]) + "\t"
report += ','.join([str(x) for x in gapsizes]) + "\t"
report += str(besttotalcov) + "\t"
report += str(bestscore) + "\t"
report += str(bestsinglescore) + "\t"
report += str(','.join(query_target_coverages) + "\t")
#if args.best_report:
# best_report_fh.write(report+"\n")
#for i in bestpath:
# args.output.write(mpa.entries[i].get_line()+"\n")
return [report, [mpa.entries[i].get_line() for i in bestpath]]
def main():
parser = argparse.ArgumentParser(
description=
"Rename gene and transcript elements of GenePred file that are redundant. Please specify an output if you would like report files generated for the filters."
)
parser.add_argument('input', help="GENEPREDFILE or '-' for STDIN")
parser.add_argument(
'-o',
'--output',
help=
"OUTPUT FILE default is STDOUT, but you need to specify an output file to get report files generated"
)
parser.add_argument(
'--minimum_locus_distance',
type=int,
default=500000,
help="Genes with the same name will be renamed if this far apart")
parser.add_argument(
'--keep_positional_duplicates',
action='store_true',
help="By default we remove one of the duplicate entries")
parser.add_argument(
'--keep_transcript_names',
action='store_true',
help="By default we rename duplicated transcript names")
parser.add_argument(
'--keep_gene_names',
action='store_true',
help="By default we rename genes located at different loci.")
args = parser.parse_args()
inf = sys.stdin
if args.input != '-': inf = open(args.input)
of = sys.stdout
if args.output: of = open(args.output, 'w')
txdef = {}
gfams = {}
for line in inf:
if line[0] == '#': continue
g = GenePredEntry(line)
loc = g.value('chrom') + ':' + ','.join(
[str(x) for x in g.value('exonStarts')]) + '-' + ','.join(
[str(x)
for x in g.value('exonEnds')]) + '/' + g.value('strand')
if loc not in txdef:
txdef[loc] = []
txdef[loc].append(g)
if g.value('gene_name') not in gfams: gfams[g.value('gene_name')] = []
gfams[g.value('gene_name')].append(g.value('name'))
# now we have cataloged all transcripts by unique locations
omissions = []
keepers = []
for loc in sorted(txdef.keys()):
if args.keep_positional_duplicates: # We don't want to ommit anything here
for g in txdef[loc]:
keepers.append(g)
continue #basically skipping this part by populating keepers
num = len(txdef[loc])
if num > 1:
sys.stderr.write("Found " + str(num) + " entries at location\n")
sys.stderr.write(loc + "\n")
sys.stderr.write("They are:\n")
largest = 0
keepgene = None
keepindex = -1
i = 0
for e in txdef[loc]:
famsize = len(gfams[e.value('gene_name')])
sys.stderr.write(" " + e.value('gene_name') + "\t" +
e.value('name') + "\t" + str(famsize) + "\n")
if famsize > largest:
keepgene = e
largest = famsize
keepindex = i
i += 1
for j in range(0, len(txdef[loc])):
if j != keepindex: omissions.append(txdef[loc][j])
else: keepers.append(txdef[loc][j])
sys.stderr.write(" Biggest gene family is " +
keepgene.value('gene_name') + " with " +
str(largest) + " transcripts\n")
sys.stderr.write(" so keep that one.\n")
else:
keepers.append(txdef[loc][0])
sys.stderr.write("Omitting " + str(len(omissions)) +
" entries for redundant positions\n")
if args.output and not args.keep_positional_duplicates:
of1 = open(args.output + '.positional_duplicate_omissions', 'w')
for g in omissions:
of1.write(g.get_line() + "\n")
of1.close()
# Now the keepers contains transcripts with unique locations
# Lets provide unique names to remaining transcripts
tnames = {}
renametx = {}
for g in keepers:
tx = g.value('name')
if tx not in tnames: tnames[tx] = []
tnames[tx].append(g)
for name in tnames:
if args.keep_transcript_names: continue # We don't want to rename them
nsize = len(tnames[name])
if nsize > 1:
sys.stderr.write("Name: " + name + " has a family of size " +
str(nsize) + "\n")
for i in range(0, len(tnames[name])):
newname = name + '[' + str(i + 1) + '/' + str(nsize) + ']'
renametx[newname] = name
tnames[name][i].entry['name'] = newname
sys.stderr.write("Renamed: " + str(len(renametx)) + " transcripts\n")
if args.output and not args.keep_transcript_names:
of1 = open(args.output + '.renamed_transcripts', 'w')
for name in sorted(renametx.keys()):
of1.write(name + "\t" + renametx[name] + "\n")
of1.close()
#now we need to arrange into gene families
gnames = {}
for name in tnames:
for g in tnames[name]:
gene = g.value('gene_name')
if gene not in gnames: gnames[gene] = []
gnames[gene].append(g)
renamegene = {}
finished = []
for gene in gnames:
if args.keep_gene_names:
for g in gnames[gene]:
finished.append(g)
continue # We don't want to rename genes
if len(gnames[gene]) == 1:
finished.append(gnames[gene][0])
continue
# Now we need to make sure these genes are really on the same locus.
loci = Loci()
loci.set_minimum_distance(args.minimum_locus_distance)
for g in gnames[gene]:
r = g.locus_range.copy()
r.set_payload(g)
loc = Locus()
loc.add_member(r)
loci.add_locus(loc)
loci.update_loci()
lcount = len(loci.loci)
if lcount == 1:
for g in gnames[gene]:
finished.append(g)
continue
# need to rename some genes
for i in range(0, lcount):
newname = gene + '[' + str(i + 1) + '/' + str(lcount) + ']'
rstr = loci.loci[i].range.get_range_string()
renamegene[newname] = gene
sys.stderr.write(newname + "\t" + rstr + "\n")
for m in loci.loci[i].members:
m.get_payload().entry['gene_name'] = newname
finished.append(m.get_payload())
sys.stderr.write("Renamed: " + str(len(renamegene)) + " genes\n")
if args.output and not args.keep_transcript_names:
of1 = open(args.output + '.renamed_genes', 'w')
for name in sorted(renamegene.keys()):
of1.write(name + "\t" + renamegene[name] + "\n")
of1.close()
#Now lets resort by genes
bygene = {}
for g in finished:
gene = g.value('gene_name')
if gene not in bygene: bygene[gene] = []
bygene[gene].append(g)
for gene in sorted(bygene.keys()):
for g in bygene[gene]:
of.write(g.get_line() + "\n")
of.close()
inf.close()
def process_read(mpa,args):
# Filter entries by a minimum alignment coverage
newentries = []
for i in [i for i in range(0,len(mpa.entries)) if mpa.entries[i].get_coverage() > args.minimum_alignment_coverage]:
newentries.append(mpa.entries[i])
mpa.entries = newentries
# Find best singles
bestsingle = None
bestsinglescore = -1
for i in range(0,len(mpa.entries)):
totalcov = mpa.entries[i].get_coverage()
weightedcov = float(mpa.entries[i].get_coverage())*float(mpa.entries[i].get_quality())
if weightedcov > bestsinglescore:
bestsinglescore = weightedcov
bestsingle = i
if bestsinglescore == -1:
sys.stderr.write("failed to find a single path\n")
return None
my_max_intron = args.maximum_intron
if args.fusion: my_max_intron = -1 # we can look any distance for a group
mpa.compatible_graph(max_intron=my_max_intron,max_query_overlap=args.maximum_query_overlap,max_gap=args.maximum_query_gap,max_target_overlap=args.maximum_target_overlap,max_query_fraction_overlap=args.maximum_query_fraction_overlap)
ps = mpa.get_root_paths()
bestpath = [bestsingle]
bestscore = 0
besttotalcov = 0
allscores = []
allcov = []
best_path_index = -1
zz = 0
for path in ps:
totalcov = sum([mpa.entries[i].get_coverage() for i in path])
weightedcov = sum([float(mpa.entries[i].get_coverage())*float(mpa.entries[i].get_quality()) for i in path])
allscores.append(weightedcov)
allcov.append(totalcov)
if weightedcov > bestscore:
bestscore = weightedcov
bestpath = path
besttotalcov = totalcov
best_path_index = zz
zz+=1
#if not bestpath: return None
otherpaths = []
for i in range(0,len(ps)):
if i != best_path_index:
otherpaths.append(ps[i])
query_target_coverages = []
for other_path in otherpaths:
qcov = 0
tcov = 0
for other_entry in [mpa.entries[i] for i in other_path]:
for entry in [mpa.entries[j] for j in bestpath]:
qcov += other_entry.query_overlap_size(entry)
tcov += other_entry.target_overlap_size(entry)
query_target_coverages.append(str(qcov)+'/'+str(tcov))
gapsizes = []
if len(bestpath) > 1:
gapsizes = [mpa.entries[bestpath[j+1]].get_query_bed().start - mpa.entries[bestpath[j]].get_query_bed().end -1 for j in range(0,len(bestpath)-1)]
#print mpa.g.get_status_string()
#print [mpa.entries[i].get_target_bed().get_range_string() for i in bestpath]
#print [mpa.entries[i].get_query_bed().get_range_string() for i in bestpath]
#print [mpa.entries[i].get_quality() for i in bestpath]
#print [mpa.entries[i].get_coverage() for i in bestpath]
#print gapsizes
#print bestscore
#print bestsinglescore
#See if we should use the single path score instead
if len(path) > 1 and bestsinglescore*(1+args.multipath_score_improvement) > bestscore:
bestpath = [bestsingle]
besttotalcov = mpa.entries[bestsingle].get_coverage()
bestscore = bestsinglescore
query_span = mpa.entries[bestpath[0]].get_query_bed()
loci = Loci()
loci.set_use_direction(True)
loci.set_minimum_distance(args.maximum_intron)
for i in bestpath:
r = mpa.entries[i].get_target_bed()
locus = Locus()
locus.set_use_direction(True)
locus.add_member(r)
loci.add_locus(locus)
loci.update_loci()
if len(bestpath) > 1:
for i in bestpath[1:]:
query_span = mpa.entries[i].get_query_bed().merge(query_span)
report = ''
report += mpa.entries[bestpath[0]].value('qName')+"\t"
report += str(len(bestpath))+"\t"
report += str(len(loci.loci))+"\t"
report += query_span.get_range_string()+"\t"
report += ','.join([mpa.entries[i].value('strand') for i in bestpath])+"\t"
report += ','.join([mpa.entries[i].get_query_bed().get_range_string() for i in bestpath])+"\t"
report += ','.join([mpa.entries[i].get_target_bed().get_range_string() for i in bestpath])+"\t"
report += ','.join([str(mpa.entries[i].get_quality()) for i in bestpath])+"\t"
report += ','.join([str(mpa.entries[i].get_coverage()) for i in bestpath])+"\t"
report += ','.join([str(x) for x in gapsizes])+"\t"
report += str(besttotalcov)+"\t"
report += str(bestscore)+"\t"
report += str(bestsinglescore)+"\t"
report += str(','.join(query_target_coverages)+"\t")
#if args.best_report:
# best_report_fh.write(report+"\n")
#for i in bestpath:
# args.output.write(mpa.entries[i].get_line()+"\n")
return [report, [mpa.entries[i].get_line() for i in bestpath]]
def main():
parser = argparse.ArgumentParser(description="Rename gene and transcript elements of GenePred file that are redundant. Please specify an output if you would like report files generated for the filters.")
parser.add_argument('input',help="GENEPREDFILE or '-' for STDIN")
parser.add_argument('-o','--output',help="OUTPUT FILE default is STDOUT, but you need to specify an output file to get report files generated")
parser.add_argument('--minimum_locus_distance',type=int,default=500000,help="Genes with the same name will be renamed if this far apart")
parser.add_argument('--keep_positional_duplicates',action='store_true',help="By default we remove one of the duplicate entries")
parser.add_argument('--keep_transcript_names',action='store_true',help="By default we rename duplicated transcript names")
parser.add_argument('--keep_gene_names',action='store_true',help="By default we rename genes located at different loci.")
args = parser.parse_args()
inf = sys.stdin
if args.input != '-': inf = open(args.input)
of = sys.stdout
if args.output: of = open(args.output,'w')
txdef = {}
gfams = {}
for line in inf:
if line[0] == '#': continue
g = GenePredEntry(line)
loc = g.value('chrom') + ':' +','.join([str(x) for x in g.value('exonStarts')]) + '-' + ','.join([str(x) for x in g.value('exonEnds')])+'/'+g.value('strand')
if loc not in txdef:
txdef[loc] = []
txdef[loc].append(g)
if g.value('gene_name') not in gfams: gfams[g.value('gene_name')] = []
gfams[g.value('gene_name')].append(g.value('name'))
# now we have cataloged all transcripts by unique locations
omissions = []
keepers = []
for loc in sorted(txdef.keys()):
if args.keep_positional_duplicates: # We don't want to ommit anything here
for g in txdef[loc]: keepers.append(g)
continue #basically skipping this part by populating keepers
num = len(txdef[loc])
if num > 1:
sys.stderr.write("Found "+str(num)+" entries at location\n")
sys.stderr.write(loc +"\n")
sys.stderr.write("They are:\n")
largest = 0
keepgene = None
keepindex = -1
i = 0
for e in txdef[loc]:
famsize = len(gfams[e.value('gene_name')])
sys.stderr.write(" "+e.value('gene_name')+"\t"+e.value('name')+"\t"+str(famsize)+"\n")
if famsize > largest:
keepgene = e
largest = famsize
keepindex = i
i+=1
for j in range(0,len(txdef[loc])):
if j != keepindex: omissions.append(txdef[loc][j])
else: keepers.append(txdef[loc][j])
sys.stderr.write(" Biggest gene family is "+keepgene.value('gene_name')+" with "+str(largest)+" transcripts\n")
sys.stderr.write(" so keep that one.\n")
else:
keepers.append(txdef[loc][0])
sys.stderr.write("Omitting "+str(len(omissions))+" entries for redundant positions\n")
if args.output and not args.keep_positional_duplicates:
of1 = open(args.output+'.positional_duplicate_omissions','w')
for g in omissions:
of1.write(g.get_line()+"\n")
of1.close()
# Now the keepers contains transcripts with unique locations
# Lets provide unique names to remaining transcripts
tnames = {}
renametx = {}
for g in keepers:
tx = g.value('name')
if tx not in tnames: tnames[tx] = []
tnames[tx].append(g)
for name in tnames:
if args.keep_transcript_names: continue # We don't want to rename them
nsize = len(tnames[name])
if nsize > 1:
sys.stderr.write("Name: "+name+" has a family of size "+str(nsize)+"\n")
for i in range(0,len(tnames[name])):
newname = name+'['+str(i+1)+'/'+str(nsize)+']'
renametx[newname] = name
tnames[name][i].entry['name'] = newname
sys.stderr.write("Renamed: "+str(len(renametx))+" transcripts\n")
if args.output and not args.keep_transcript_names:
of1 = open(args.output+'.renamed_transcripts','w')
for name in sorted(renametx.keys()):
of1.write(name+"\t"+renametx[name]+"\n")
of1.close()
#now we need to arrange into gene families
gnames = {}
for name in tnames:
for g in tnames[name]:
gene = g.value('gene_name')
if gene not in gnames: gnames[gene] = []
gnames[gene].append(g)
renamegene = {}
finished = []
for gene in gnames:
if args.keep_gene_names:
for g in gnames[gene]: finished.append(g)
continue # We don't want to rename genes
if len(gnames[gene])==1:
finished.append(gnames[gene][0])
continue
# Now we need to make sure these genes are really on the same locus.
loci = Loci()
loci.set_minimum_distance(args.minimum_locus_distance)
for g in gnames[gene]:
r = g.locus_range.copy()
r.set_payload(g)
loc = Locus()
loc.add_member(r)
loci.add_locus(loc)
loci.update_loci()
lcount = len(loci.loci)
if lcount == 1:
for g in gnames[gene]: finished.append(g)
continue
# need to rename some genes
for i in range(0,lcount):
newname = gene+'['+str(i+1)+'/'+str(lcount)+']'
rstr = loci.loci[i].range.get_range_string()
renamegene[newname] = gene
sys.stderr.write(newname+"\t"+rstr+"\n")
for m in loci.loci[i].members:
m.get_payload().entry['gene_name'] = newname
finished.append(m.get_payload())
sys.stderr.write("Renamed: "+str(len(renamegene))+" genes\n")
if args.output and not args.keep_transcript_names:
of1 = open(args.output+'.renamed_genes','w')
for name in sorted(renamegene.keys()):
of1.write(name+"\t"+renamegene[name]+"\n")
of1.close()
#Now lets resort by genes
bygene = {}
for g in finished:
gene = g.value('gene_name')
if gene not in bygene: bygene[gene] = []
bygene[gene].append(g)
for gene in sorted(bygene.keys()):
for g in bygene[gene]:
of.write(g.get_line()+"\n")
of.close()
inf.close()