def choose_best(lst):
hits = []
for m in lst:
total_introns = m.num_introns + m.ref_num_introns
if total_introns == 0:
intron_frac = 0.0
else:
intron_frac = float(
m.shared_introns) / (total_introns - m.shared_introns)
same_strand_frac = float(m.shared_same_strand_bp) / (
m.length + m.ref_length - m.shared_same_strand_bp)
opp_strand_frac = float(m.shared_opp_strand_bp) / (
m.length + m.ref_length - m.shared_opp_strand_bp)
category_int = Category.to_int(m.category)
hits.append(
(int(m.shared_splicing), intron_frac, same_strand_frac,
opp_strand_frac,
int(category_int == Category.INTRONIC_SAME_STRAND),
int(category_int == Category.INTRONIC_OPP_STRAND),
int(category_int == Category.INTERLEAVING_SAME_STRAND),
int(category_int == Category.INTERLEAVING_OPP_STRAND),
int(category_int == Category.ENCOMPASSING_SAME_STRAND),
int(category_int == Category.ENCOMPASSING_OPP_STRAND),
int(category_int == Category.INTERGENIC), -abs(m.distance),
m))
# sort matches
hits.sort(reverse=True)
hit = hits[0][-1]
return hit
def choose_best(lst):
hits = []
for m in lst:
total_introns = m.num_introns + m.ref_num_introns
if total_introns == 0:
intron_frac = 0.0
else:
intron_frac = float(m.shared_introns) / (total_introns - m.shared_introns)
same_strand_frac = float(m.shared_same_strand_bp) / (m.length + m.ref_length - m.shared_same_strand_bp)
opp_strand_frac = float(m.shared_opp_strand_bp) / (m.length + m.ref_length - m.shared_opp_strand_bp)
category_int = Category.to_int(m.category)
hits.append((int(m.shared_splicing), intron_frac,
same_strand_frac, opp_strand_frac,
int(category_int == Category.INTRONIC_SAME_STRAND),
int(category_int == Category.INTRONIC_OPP_STRAND),
int(category_int == Category.INTERLEAVING_SAME_STRAND),
int(category_int == Category.INTERLEAVING_OPP_STRAND),
int(category_int == Category.ENCOMPASSING_SAME_STRAND),
int(category_int == Category.ENCOMPASSING_OPP_STRAND),
int(category_int == Category.INTERGENIC),
-abs(m.distance), m))
# sort matches
hits.sort(reverse=True)
hit = hits[0][-1]
return hit
def impute_transcript(t, gene_map, transcript_map):
catstr = t.attrs['category']
catint = Category.to_int(catstr)
length = t.length
gene_type = t.attrs.get('gene_type', None)
ref_gene_type = t.attrs['ref_gene_type']
# ref_gene_type can be be multiple gene types separated by commas.
# convert into a set of unique gene types
ref_gene_types = set(ref_gene_type.split(','))
transcript_types = set(
impute_transcript_type(catint, length, gene_type, x)
for x in ref_gene_types)
transcript_categories = set(GENCODE_CATEGORY_MAP[x]
for x in transcript_types)
# sorted and join unique types/categories to make conglomerated category assignments
transcript_type = ','.join(sorted(transcript_types))
transcript_category = ','.join(sorted(transcript_categories))
# use first gene in read-through for name
#ref_gene_name = t.attrs['ref_gene_name'].split(',')[0]
# hyphenate read-through genes into long name
ref_gene_name = '-'.join(t.attrs['ref_gene_name'].split(','))
# resolve upper/lower case issue with gene names from
# different databases
transcript_name = ref_gene_name.upper()
# build transcript name
if transcript_name == 'NONE':
transcript_name = str(t.chrom)
# append category
if catint != Category.SAME_STRAND:
transcript_name = '%s.%s' % (transcript_name, catstr)
# transcript name string is key to a dictionary that
# associates each gene id with an integer number
gene_id = t.attrs['gene_id']
gene_dict = gene_map[transcript_name]
if gene_id not in gene_dict:
gene_num = len(gene_dict) + 1
gene_dict[gene_id] = gene_num
else:
gene_num = gene_dict[gene_id]
# append gene integers to name
transcript_name = '%s.%d' % (transcript_name, gene_num)
# gene id is also key to dict that associates each isoform
# of gene with integer number
t_id = t.attrs['transcript_id']
t_dict = transcript_map[transcript_name]
if t_id not in t_dict:
t_num = len(t_dict) + 1
t_dict[t_id] = t_num
else:
t_num = t_dict[t_id]
# append gene/transcript integers to gene name
transcript_name = '%s.%d' % (transcript_name, t_num)
return transcript_type, transcript_category, transcript_name
def consensus(lst):
if len(lst) == 0:
return None
# first check for read through transcripts involving multiple
# reference genes
same_strand_hits = collections.defaultdict(lambda: [])
for m in lst:
category_int = Category.to_int(m.category)
if category_int == Category.SAME_STRAND:
same_strand_hits[m.ref_gene_id].append(m)
# no same strand matches so don't need to worry about
# read-throughs or multiple gene types
if len(same_strand_hits) == 0:
return MatchStats.choose_best(lst)
# get consensus match from same strand overlapping genes
total_introns = lst[0].num_introns
total_length = lst[0].length
shared_introns = 0
shared_same_strand_bp = 0
hits = []
for genelst in same_strand_hits.itervalues():
m = MatchStats.choose_best(genelst).copy()
m.ref_gene_type = ','.join(
sorted(set(m.ref_gene_type for m in genelst)))
total_introns += m.ref_num_introns
total_length += m.ref_length
shared_introns += m.shared_introns
shared_same_strand_bp += m.shared_same_strand_bp
hits.append(m)
# sort reference genes by position
hits = MatchStats.sort_genome(hits)
# make a new MatchStats object
hit = hits[0].copy()
hit.ref_transcript_id = ','.join(x.ref_transcript_id for x in hits)
hit.ref_gene_id = ','.join(x.ref_gene_id for x in hits)
hit.ref_orig_gene_id = ','.join(x.ref_orig_gene_id for x in hits)
hit.ref_gene_name = ','.join(x.ref_gene_name for x in hits)
hit.ref_source = ','.join(x.ref_source for x in hits)
hit.ref_gene_type = ','.join(x.ref_gene_type for x in hits)
hit.ref_locus = ','.join(x.ref_locus for x in hits)
hit.ref_length = ','.join(str(x.ref_length) for x in hits)
hit.ref_num_introns = ','.join(str(x.ref_num_introns) for x in hits)
hit.shared_same_strand_bp = shared_same_strand_bp
hit.shared_opp_strand_bp = 0
hit.shared_introns = shared_introns
hit.shared_splicing = any(m.shared_splicing for m in hits)
hit.distance = 0
if len(same_strand_hits) > 1:
hit.category = Category.to_str(Category.READ_THROUGH)
return hit
def impute_transcript(t, gene_map, transcript_map):
catstr = t.attrs['category']
catint = Category.to_int(catstr)
length = t.length
gene_type = t.attrs.get('gene_type', None)
ref_gene_type = t.attrs['ref_gene_type']
# ref_gene_type can be be multiple gene types separated by commas.
# convert into a set of unique gene types
ref_gene_types = set(ref_gene_type.split(','))
transcript_types = set(impute_transcript_type(catint, length, gene_type, x) for x in ref_gene_types)
transcript_categories = set(GENCODE_CATEGORY_MAP[x] for x in transcript_types)
# sorted and join unique types/categories to make conglomerated category assignments
transcript_type = ','.join(sorted(transcript_types))
transcript_category = ','.join(sorted(transcript_categories))
# use first gene in read-through for name
#ref_gene_name = t.attrs['ref_gene_name'].split(',')[0]
# hyphenate read-through genes into long name
ref_gene_name = '-'.join(t.attrs['ref_gene_name'].split(','))
# resolve upper/lower case issue with gene names from
# different databases
transcript_name = ref_gene_name.upper()
# build transcript name
if transcript_name == 'NONE':
transcript_name = str(t.chrom)
# append category
if catint != Category.SAME_STRAND:
transcript_name = '%s.%s' % (transcript_name, catstr)
# transcript name string is key to a dictionary that
# associates each gene id with an integer number
gene_id = t.attrs['gene_id']
gene_dict = gene_map[transcript_name]
if gene_id not in gene_dict:
gene_num = len(gene_dict) + 1
gene_dict[gene_id] = gene_num
else:
gene_num = gene_dict[gene_id]
# append gene integers to name
transcript_name = '%s.%d' % (transcript_name, gene_num)
# gene id is also key to dict that associates each isoform
# of gene with integer number
t_id = t.attrs['transcript_id']
t_dict = transcript_map[transcript_name]
if t_id not in t_dict:
t_num = len(t_dict) + 1
t_dict[t_id] = t_num
else:
t_num = t_dict[t_id]
# append gene/transcript integers to gene name
transcript_name = '%s.%d' % (transcript_name, t_num)
return transcript_type, transcript_category, transcript_name
def consensus(lst):
if len(lst) == 0:
return None
# first check for read through transcripts involving multiple
# reference genes
same_strand_hits = collections.defaultdict(lambda: [])
for m in lst:
category_int = Category.to_int(m.category)
if category_int == Category.SAME_STRAND:
same_strand_hits[m.ref_gene_id].append(m)
# no same strand matches so don't need to worry about
# read-throughs or multiple gene types
if len(same_strand_hits) == 0:
return MatchStats.choose_best(lst)
# get consensus match from same strand overlapping genes
total_introns = lst[0].num_introns
total_length = lst[0].length
shared_introns = 0
shared_same_strand_bp = 0
hits = []
for genelst in same_strand_hits.itervalues():
m = MatchStats.choose_best(genelst).copy()
m.ref_gene_type = ','.join(sorted(set(m.ref_gene_type for m in genelst)))
total_introns += m.ref_num_introns
total_length += m.ref_length
shared_introns += m.shared_introns
shared_same_strand_bp += m.shared_same_strand_bp
hits.append(m)
# sort reference genes by position
hits = MatchStats.sort_genome(hits)
# make a new MatchStats object
hit = hits[0].copy()
hit.ref_transcript_id = ','.join(x.ref_transcript_id for x in hits)
hit.ref_gene_id = ','.join(x.ref_gene_id for x in hits)
hit.ref_orig_gene_id = ','.join(x.ref_orig_gene_id for x in hits)
hit.ref_gene_name = ','.join(x.ref_gene_name for x in hits)
hit.ref_source = ','.join(x.ref_source for x in hits)
hit.ref_gene_type = ','.join(x.ref_gene_type for x in hits)
hit.ref_locus = ','.join(x.ref_locus for x in hits)
hit.ref_length = ','.join(str(x.ref_length) for x in hits)
hit.ref_num_introns = ','.join(str(x.ref_num_introns) for x in hits)
hit.shared_same_strand_bp = shared_same_strand_bp
hit.shared_opp_strand_bp = 0
hit.shared_introns = shared_introns
hit.shared_splicing = any(m.shared_splicing for m in hits)
hit.distance = 0
if len(same_strand_hits) > 1:
hit.category = Category.to_str(Category.READ_THROUGH)
return hit
def main():
logging.basicConfig(level=logging.DEBUG,
format="%(asctime)s - %(name)s - %(levelname)s - %(message)s")
parser = argparse.ArgumentParser()
parser.add_argument('--rename', dest='rename', action='store_true')
parser.add_argument('gtf_file')
args = parser.parse_args()
gtf_file = args.gtf_file
rename = args.rename
if not os.path.exists(gtf_file):
parser.error("GTF file '%s' not found" % (gtf_file))
# parse transcripts
num_transcripts = 0
# keep track of redundant gene/transcript counts
gene_map = collections.defaultdict(lambda: {})
transcript_map = collections.defaultdict(lambda: {})
for transcripts in parse_gtf(open(gtf_file)):
for t in transcripts:
catstr = t.attrs['category']
catint = Category.to_int(catstr)
gene_type = t.attrs.get('gene_type', None)
ref_gene_type = t.attrs['ref_gene_type']
if catint == Category.SAME_STRAND:
# impute gene type
new_gene_type = ref_gene_type
else:
if gene_type == 'protein_coding':
# don't change protein coding genes
new_gene_type = gene_type
elif t.length < 250:
# categorize small RNA separately
new_gene_type = 'misc_RNA'
else:
if ref_gene_type == 'protein_coding':
# categorize based on overlap with reference
new_gene_type = PROTEIN_CATEGORY_MAP[catint]
else:
# reference is also non-coding
new_gene_type = 'lincRNA'
# get gene category
gene_category = GENCODE_CATEGORY_MAP[new_gene_type]
new_gene_name = None
if rename:
# resolve upper/lower case issue with gene names from
# different databases
ref_gene_name = t.attrs['ref_gene_name'].upper()
# build new gene name
if ref_gene_name == 'NONE':
new_gene_name = str(t.attrs['source'])
elif catint == Category.SAME_STRAND:
new_gene_name = str(ref_gene_name)
else:
new_gene_name = '%s.%s' % (ref_gene_name, catstr)
# gene name string is key to a dictionary that
# associates each gene id with an integer number
gene_id = t.attrs['gene_id']
gene_dict = gene_map[new_gene_name]
if gene_id not in gene_dict:
gene_num = len(gene_dict) + 1
gene_dict[gene_id] = gene_num
else:
gene_num = gene_dict[gene_id]
# gene id is also key to dict that associates each isoform
# of gene with integer number
t_id = t.attrs['transcript_id']
t_dict = transcript_map[gene_id]
if t_id not in t_dict:
t_num = len(t_dict) + 1
t_dict[t_id] = t_num
else:
t_num = t_dict[t_id]
# append gene/transcript integers to gene name
new_gene_name = '%s.%d.%d' % (new_gene_name, gene_num, t_num)
# write new attributes
for f in t.to_gtf_features(source='assemblyline', score=1000):
f.attrs['gene_type'] = new_gene_type
f.attrs['gene_category'] = gene_category
if rename:
if 'gene_name' in f.attrs:
f.attrs['orig_gene_name'] = f.attrs['gene_name']
f.attrs['gene_name'] = new_gene_name
print str(f)
num_transcripts += 1
return 0
def main():
logging.basicConfig(
level=logging.DEBUG,
format="%(asctime)s - %(name)s - %(levelname)s - %(message)s")
parser = argparse.ArgumentParser()
parser.add_argument('--rename', dest='rename', action='store_true')
parser.add_argument('gtf_file')
args = parser.parse_args()
gtf_file = args.gtf_file
rename = args.rename
if not os.path.exists(gtf_file):
parser.error("GTF file '%s' not found" % (gtf_file))
# parse transcripts
num_transcripts = 0
# keep track of redundant gene/transcript counts
gene_map = collections.defaultdict(lambda: {})
transcript_map = collections.defaultdict(lambda: {})
for transcripts in parse_gtf(open(gtf_file)):
for t in transcripts:
catstr = t.attrs['category']
catint = Category.to_int(catstr)
gene_type = t.attrs.get('gene_type', None)
ref_gene_type = t.attrs['ref_gene_type']
if catint == Category.SAME_STRAND:
# impute gene type
new_gene_type = ref_gene_type
else:
if gene_type == 'protein_coding':
# don't change protein coding genes
new_gene_type = gene_type
elif t.length < 250:
# categorize small RNA separately
new_gene_type = 'misc_RNA'
else:
if ref_gene_type == 'protein_coding':
# categorize based on overlap with reference
new_gene_type = PROTEIN_CATEGORY_MAP[catint]
else:
# reference is also non-coding
new_gene_type = 'lincRNA'
# get gene category
gene_category = GENCODE_CATEGORY_MAP[new_gene_type]
new_gene_name = None
if rename:
# resolve upper/lower case issue with gene names from
# different databases
ref_gene_name = t.attrs['ref_gene_name'].upper()
# build new gene name
if ref_gene_name == 'NONE':
new_gene_name = str(t.attrs['source'])
elif catint == Category.SAME_STRAND:
new_gene_name = str(ref_gene_name)
else:
new_gene_name = '%s.%s' % (ref_gene_name, catstr)
# gene name string is key to a dictionary that
# associates each gene id with an integer number
gene_id = t.attrs['gene_id']
gene_dict = gene_map[new_gene_name]
if gene_id not in gene_dict:
gene_num = len(gene_dict) + 1
gene_dict[gene_id] = gene_num
else:
gene_num = gene_dict[gene_id]
# gene id is also key to dict that associates each isoform
# of gene with integer number
t_id = t.attrs['transcript_id']
t_dict = transcript_map[gene_id]
if t_id not in t_dict:
t_num = len(t_dict) + 1
t_dict[t_id] = t_num
else:
t_num = t_dict[t_id]
# append gene/transcript integers to gene name
new_gene_name = '%s.%d.%d' % (new_gene_name, gene_num, t_num)
# write new attributes
for f in t.to_gtf_features(source='assemblyline', score=1000):
f.attrs['gene_type'] = new_gene_type
f.attrs['gene_category'] = gene_category
if rename:
if 'gene_name' in f.attrs:
f.attrs['orig_gene_name'] = f.attrs['gene_name']
f.attrs['gene_name'] = new_gene_name
print str(f)
num_transcripts += 1
return 0
