def parse_gene_chimeric_reads(bamfh):
# create a dictionary structure to hold read pairs
chimera_dict = collections.defaultdict(lambda: [])
for r1, r2 in parse_pairs(bamfh):
#
# TODO:
# for now we are only going to deal with gene-gene
# chimeras and leave other chimeras for study at a
# later time
#
dr1 = r1.opt(DISCORDANT_TAG_NAME)
dr2 = r2.opt(DISCORDANT_TAG_NAME)
if (dr1 != DiscordantTags.DISCORDANT_GENE
or dr2 != DiscordantTags.DISCORDANT_GENE):
continue
# organize key in 5' to 3' order
or1 = r1.opt(ORIENTATION_TAG_NAME)
or2 = r2.opt(ORIENTATION_TAG_NAME)
assert or1 != or2
if or1 == OrientationTags.FIVEPRIME:
pair = (r1, r2)
else:
pair = (r2, r1)
# store pertinent information in lightweight structure
# convert to DiscordantRead objects
r5p = DiscordantRead.from_read(pair[0])
r3p = DiscordantRead.from_read(pair[1])
# keep list of discordant chimeric reads
chimera_dict[(r5p.tid, r3p.tid)].append((r5p, r3p))
for key, pairs in chimera_dict.iteritems():
rname1, rname2 = key
yield rname1, rname2, pairs
def parse_gene_chimeric_reads(bamfh):
# create a dictionary structure to hold read pairs
chimera_dict = collections.defaultdict(lambda: [])
for r1,r2 in parse_pairs(bamfh):
#
# TODO:
# for now we are only going to deal with gene-gene
# chimeras and leave other chimeras for study at a
# later time
#
dr1 = r1.opt(DISCORDANT_TAG_NAME)
dr2 = r2.opt(DISCORDANT_TAG_NAME)
if (dr1 != DiscordantTags.DISCORDANT_GENE or
dr2 != DiscordantTags.DISCORDANT_GENE):
continue
# organize key in 5' to 3' order
or1 = r1.opt(ORIENTATION_TAG_NAME)
or2 = r2.opt(ORIENTATION_TAG_NAME)
assert or1 != or2
if or1 == OrientationTags.FIVEPRIME:
pair = (r1,r2)
else:
pair = (r2,r1)
# store pertinent information in lightweight structure
# convert to DiscordantRead objects
r5p = DiscordantRead.from_read(pair[0])
r3p = DiscordantRead.from_read(pair[1])
# keep list of discordant chimeric reads
chimera_dict[(r5p.tid, r3p.tid)].append((r5p,r3p))
for key,pairs in chimera_dict.iteritems():
rname1,rname2 = key
yield rname1, rname2, pairs
def discordant_reads_to_bedpe(index_dir, input_bam_file, output_file):
# open BAM alignment file
bamfh = pysam.Samfile(input_bam_file, "rb")
# build a lookup table to get genomic intervals from transcripts
logging.debug("Reading gene information")
gene_file = os.path.join(index_dir, config.GENE_FEATURE_FILE)
tid_gene_map = build_tid_gene_map(bamfh, gene_file,
rname_prefix=config.GENE_REF_PREFIX)
outfh = open(output_file, "w")
logging.debug("Converting BAM to BEDPE format")
for r5p,r3p in parse_gene_discordant_reads(bamfh):
# store pertinent read information in lightweight structure called
# DiscordantRead object. this departs from SAM format into a
# custom read format
dr5p = DiscordantRead.from_read(r5p)
dr3p = DiscordantRead.from_read(r3p)
# get gene information
tx5p = tid_gene_map[r5p.rname]
tx3p = tid_gene_map[r3p.rname]
# write bedpe format
fields = [tx5p.tx_name, r5p.pos, r5p.aend,
tx3p.tx_name, r3p.pos, r3p.aend,
r5p.qname, # read name
0, # score
tx5p.strand, tx3p.strand, # strand 1, strand 2
]
fields.append('|'.join(map(str, dr5p.to_list())))
fields.append('|'.join(map(str, dr3p.to_list())))
print >>outfh, '\t'.join(map(str, fields))
outfh.close()
def discordant_reads_to_bedpe(index_dir, input_bam_file, output_file):
# open BAM alignment file
bamfh = pysam.Samfile(input_bam_file, "rb")
# build a lookup table to get genomic intervals from transcripts
logging.debug("Reading transcript features")
transcript_file = os.path.join(index_dir, config.TRANSCRIPT_FEATURE_FILE)
transcripts = list(TranscriptFeature.parse(open(transcript_file)))
tid_tx_map = build_tid_transcript_map(bamfh, transcripts)
outfh = open(output_file, "w")
logging.debug("Converting BAM to BEDPE format")
for r5p,r3p in parse_gene_discordant_reads(bamfh):
# store pertinent read information in lightweight structure called
# DiscordantRead object. this departs from SAM format into a
# custom read format
dr5p = DiscordantRead.from_read(r5p)
dr3p = DiscordantRead.from_read(r3p)
# get gene information
tx5p = tid_tx_map[r5p.rname]
tx3p = tid_tx_map[r3p.rname]
# write bedpe format
fields = [tx5p.tx_id, r5p.pos, r5p.aend,
tx3p.tx_id, r3p.pos, r3p.aend,
r5p.qname, # read name
0, # score
tx5p.strand, tx3p.strand, # strand 1, strand 2
]
fields.append('|'.join(map(str, dr5p.to_list())))
fields.append('|'.join(map(str, dr3p.to_list())))
print >>outfh, '\t'.join(map(str, fields))
outfh.close()
bamfh.close()
def filter_spanning_reads(chimeras, reads,
anchor_min,
anchor_length,
anchor_mismatches,
library_type):
for i,r in enumerate(reads):
if r.is_unmapped:
continue
# make a discordant read object
# TODO: need to annotate reads elsewhere since they have already been sorted here
r.tags = r.tags + [("HI", 0),
("IH", 1),
("NH", 1),
(DISCORDANT_TAG_NAME, DiscordantTags.DISCORDANT_GENE),
(ORIENTATION_TAG_NAME, get_gene_orientation(r, library_type))]
dr = DiscordantRead.from_read(r)
dr.is_spanning = True
# check read alignment against chimeras
for c in chimeras:
if check_breakpoint_alignment(c, r,
anchor_min,
anchor_length,
anchor_mismatches):
# valid spanning read
yield c,dr
def parse_discordant_bedpe_by_transcript_pair(fh):
prev_tx5p, prev_tx3p = None,None
frags = []
for line in fh:
fields = line.strip().split('\t')
tx5p = fields[0]
tx3p = fields[3]
dr5p = DiscordantRead.from_list(fields[10].split("|"))
dr3p = DiscordantRead.from_list(fields[11].split("|"))
if (tx5p, tx3p) != (prev_tx5p, prev_tx3p):
if len(frags) > 0:
yield prev_tx5p, prev_tx3p, frags
frags = []
prev_tx5p, prev_tx3p = tx5p, tx3p
frags.append((dr5p, dr3p))
if len(frags) > 0:
yield tx5p, tx3p, frags
def parse_discordant_bedpe_by_transcript_pair(fh):
prev_tx5p, prev_tx3p = None, None
frags = []
for line in fh:
fields = line.strip().split('\t')
tx5p = fields[0]
tx3p = fields[3]
dr5p = DiscordantRead.from_list(fields[10].split("|"))
dr3p = DiscordantRead.from_list(fields[11].split("|"))
if (tx5p, tx3p) != (prev_tx5p, prev_tx3p):
if len(frags) > 0:
yield prev_tx5p, prev_tx3p, frags
frags = []
prev_tx5p, prev_tx3p = tx5p, tx3p
frags.append((dr5p, dr3p))
if len(frags) > 0:
yield tx5p, tx3p, frags
def filter_spanning_reads(chimeras, reads, anchor_min, anchor_length,
anchor_mismatches, library_type):
for i, r in enumerate(reads):
if r.is_unmapped:
continue
# make a discordant read object
# TODO: need to annotate reads elsewhere since they have already been sorted here
r.tags = r.tags + [
("HI", 0), ("IH", 1), ("NH", 1),
(DISCORDANT_TAG_NAME, DiscordantTags.DISCORDANT_GENE),
(ORIENTATION_TAG_NAME, get_orientation(r, library_type))
]
dr = DiscordantRead.from_read(r)
dr.is_spanning = True
# check read alignment against chimeras
for c in chimeras:
if check_breakpoint_alignment(c, r, anchor_min, anchor_length,
anchor_mismatches):
# valid spanning read
yield c, dr
def discordant_reads_to_breakpoints(index_dir, isize_dist_file, input_bam_file,
output_file, trim_bp, max_read_length,
homology_mismatches):
"""
homology_mismatches: number of mismatches to tolerate while computing
homology between chimeric breakpoint sequence and "wildtype" sequence
trim_bp: when selecting the best matching exon for each read, we
account for spurious overlap into adjacent exons by trimming the
read by 'trim_bp'
"""
# read insert size distribution
isize_dist = InsertSizeDistribution.from_file(open(isize_dist_file))
# open BAM alignment file
bamfh = pysam.Samfile(input_bam_file, "rb")
# build a lookup table to get genomic intervals from transcripts
logging.debug("Reading gene information")
gene_file = os.path.join(index_dir, config.GENE_FEATURE_FILE)
tid_tx_map = build_tid_tx_map(bamfh,
gene_file,
rname_prefix=config.GENE_REF_PREFIX)
# open the reference sequence fasta file
ref_fasta_file = os.path.join(index_dir, config.ALIGN_INDEX + ".fa")
ref_fa = pysam.Fastafile(ref_fasta_file)
# iterate through read pairs
outfh = open(output_file, "w")
logging.debug("Parsing discordant reads")
for r5p, r3p in parse_gene_discordant_reads(bamfh):
# store pertinent read information in lightweight structure called
# DiscordantRead object. this departs from SAM format into a
# custom read format
dr5p = DiscordantRead.from_read(r5p)
dr3p = DiscordantRead.from_read(r3p)
# get gene information
tx5p = tid_tx_map[r5p.rname]
tx3p = tid_tx_map[r3p.rname]
# given the insert size find the highest probability
# exon junction breakpoint between the two transcripts
isize_prob, breakpoints = \
choose_best_breakpoints(r5p, r3p, tx5p, tx3p,
trim_bp, isize_dist)
# extract the sequence of the breakpoint along with the
# number of homologous bases at the breakpoint between
# chimera and wildtype genes
for breakpoint in breakpoints:
exon_num_5p, tx_end_5p, exon_num_3p, tx_start_3p = breakpoint
breakpoint_seq_5p, breakpoint_seq_3p, homology_left, homology_right = \
extract_breakpoint_sequence(config.GENE_REF_PREFIX + tx5p.tx_name, tx_end_5p,
config.GENE_REF_PREFIX + tx3p.tx_name, tx_start_3p,
ref_fa, max_read_length,
homology_mismatches)
# write breakpoint information for each read to a file
fields = [
tx5p.tx_name,
0,
tx_end_5p,
tx3p.tx_name,
tx_start_3p,
tx3p.tx_end,
r5p.rname, # name
isize_prob, # score
tx5p.strand,
tx3p.strand, # strand 1, strand 2
# user defined fields
exon_num_5p,
exon_num_3p,
breakpoint_seq_5p,
breakpoint_seq_3p,
homology_left,
homology_right
]
fields.append('|'.join(map(str, dr5p.to_list())))
fields.append('|'.join(map(str, dr3p.to_list())))
print >> outfh, '\t'.join(map(str, fields))
# cleanup
ref_fa.close()
outfh.close()
bamfh.close()
return config.JOB_SUCCESS
def discordant_reads_to_breakpoints(index_dir, isize_dist_file,
input_bam_file, output_file,
trim_bp, max_read_length,
homology_mismatches):
"""
homology_mismatches: number of mismatches to tolerate while computing
homology between chimeric breakpoint sequence and "wildtype" sequence
trim_bp: when selecting the best matching exon for each read, we
account for spurious overlap into adjacent exons by trimming the
read by 'trim_bp'
"""
# read insert size distribution
isize_dist = InsertSizeDistribution.from_file(open(isize_dist_file))
# open BAM alignment file
bamfh = pysam.Samfile(input_bam_file, "rb")
# build a lookup table to get genomic intervals from transcripts
logging.debug("Reading gene information")
gene_file = os.path.join(index_dir, config.GENE_FEATURE_FILE)
tid_tx_map = build_tid_tx_map(bamfh, gene_file,
rname_prefix=config.GENE_REF_PREFIX)
# open the reference sequence fasta file
ref_fasta_file = os.path.join(index_dir, config.ALIGN_INDEX + ".fa")
ref_fa = pysam.Fastafile(ref_fasta_file)
# iterate through read pairs
outfh = open(output_file, "w")
logging.debug("Parsing discordant reads")
for r5p,r3p in parse_gene_discordant_reads(bamfh):
# store pertinent read information in lightweight structure called
# DiscordantRead object. this departs from SAM format into a
# custom read format
dr5p = DiscordantRead.from_read(r5p)
dr3p = DiscordantRead.from_read(r3p)
# get gene information
tx5p = tid_tx_map[r5p.rname]
tx3p = tid_tx_map[r3p.rname]
# given the insert size find the highest probability
# exon junction breakpoint between the two transcripts
isize_prob, breakpoints = \
choose_best_breakpoints(r5p, r3p, tx5p, tx3p,
trim_bp, isize_dist)
# extract the sequence of the breakpoint along with the
# number of homologous bases at the breakpoint between
# chimera and wildtype genes
for breakpoint in breakpoints:
exon_num_5p, tx_end_5p, exon_num_3p, tx_start_3p = breakpoint
breakpoint_seq_5p, breakpoint_seq_3p, homology_left, homology_right = \
extract_breakpoint_sequence(config.GENE_REF_PREFIX + tx5p.tx_name, tx_end_5p,
config.GENE_REF_PREFIX + tx3p.tx_name, tx_start_3p,
ref_fa, max_read_length,
homology_mismatches)
# write breakpoint information for each read to a file
fields = [tx5p.tx_name, 0, tx_end_5p,
tx3p.tx_name, tx_start_3p, tx3p.tx_end,
r5p.rname, # name
isize_prob, # score
tx5p.strand, tx3p.strand, # strand 1, strand 2
# user defined fields
exon_num_5p, exon_num_3p,
breakpoint_seq_5p, breakpoint_seq_3p,
homology_left, homology_right]
fields.append('|'.join(map(str, dr5p.to_list())))
fields.append('|'.join(map(str, dr3p.to_list())))
print >>outfh, '\t'.join(map(str, fields))
# cleanup
ref_fa.close()
outfh.close()
bamfh.close()
return config.JOB_SUCCESS
