def variant_annotate(vcf_path,
genome='~/tools/annovar-2016-02-01/humandb/hg38'):
format_annovar(vcf_path, 'anno_tmp.vcf')
humandb_dir, genome_version = os.path.split(genome)
shell('table_annovar.pl anno_tmp.vcf %s -buildver %s --remove --otherinfo '
'--outfile annotated -operation g,f,f,f '
'-protocol refGene,cosmic70,1000g2014oct_all,exac03' %
(humandb_dir, genome_version))
anno = open('annotated.%s_multianno.txt' % genome_version)
out = zopen('annotated.vcf.gz', 'w')
anno.next()
line = anno.next()
headers = [
'CHROM', 'POSITION', 'REF', 'ALT', 'FUNCTION', 'GENE',
'EXONIC_FUNCTION', 'AA_CHANGE', 'COSMIC', '1000G', 'EXAC'
]
headers += line.rstrip('\n').split('\t')[20:]
out.write('\t'.join(headers) + '\n')
for line in anno:
c = line.rstrip('\n').split('\t')
out.write('\t'.join(c[0:2] + c[3:7] + c[8:13] + c[20:]))
out.write('\n')
out.close()
os.remove('anno_tmp.vcf')
os.remove('annotated.%s_multianno.txt' % genome_version)
if num_lines('annotated.invalid_input') <= 1:
os.remove('annotated.invalid_input')
if num_lines('annotated.refGene.invalid_input') <= 1:
os.remove('annotated.refGene.invalid_input')
def sam_count(bam_path,
bed_path,
genome_path='~/organisms/homo_sapiens/hg19.chrom.sizes'):
bed_cols = len(next(open(bed_path)).split('\t'))
shell(
'bedtools coverage -split -sorted -counts -g %s -a %s -b %s | cut -f%d'
% (genome_path, bed_path, bam_path, bed_cols + 1))
def swiss_download_sra(sra_study):
if not sra_study.startswith('SRP'):
error('SRA study identifier must begin with "SRP".')
shell('/data/csb/tools/ncftp-3.2.5/bin/ncftpget -R -v '
'ftp-trace.ncbi.nlm.nih.gov ./ '
'/sra/sra-instant/reads/ByStudy/sra/SRP/%s/%s' %
(sra_study[:6], sra_study))
def lftp_mirror(rule, dry_run=False):
cmds = open('.lftp_script', 'w')
cmds.write('open -u %s,%s sftp://%s\n' % (
rule.username, rule.password, rule.dst_host))
cmds.write('mirror -P3 -Rae %s %s %s\n' % (
'--dry-run' if dry_run else '-v', rule.src_dir, rule.dst_dir))
cmds.close()
if dry_run:
userpass = rule.username + ':' + rule.password + '@'
host = rule.dst_host
out = shell_stdout('lftp -f .lftp_script')
for line in out:
if line.startswith('chmod'): continue
if line.startswith('mkdir'): continue
m = re.match('get -O sftp://(.+) file:/.+/(.+)', line)
if m:
dst = m.group(1)
if dst.startswith(userpass): dst = dst[len(userpass):]
if dst.startswith(host): dst = dst[len(host):]
print('ADD %s/%s' % (dst, m.group(2)))
continue
m = re.match('get -e -O sftp://(.+) file:/.+/(.+)', line)
if m:
dst = m.group(1)
if dst.startswith(userpass): dst = dst[len(userpass):]
if dst.startswith(host): dst = dst[len(host):]
print('UPDATE %s/%s' % (dst, m.group(2)))
continue
m = re.match('rm .*sftp://(.+)', line)
if m:
dst = m.group(1)
if dst.startswith(userpass): dst = dst[len(userpass):]
if dst.startswith(host): dst = dst[len(host):]
print('DELETE %s' % dst)
continue
sys.stdout.write(line)
else:
shell('lftp -f .lftp_script')
os.remove('.lftp_script')
def sam_compact(bam_path):
shell('samtools view -H %s' % bam_path)
short_id = {}
id_counter = 1
for al in read_sam(bam_path, 'D'):
rid = al[0]
if rid[-2] == '/': rid = rid[:-2]
new_id = short_id.get(rid)
if new_id:
del short_id[rid]
else:
new_id = str(id_counter)
id_counter += 1
short_id[rid] = new_id
al[0] = new_id
al[10] = '*'
sys.stdout.write('\t'.join(al))
def parallel_worker(log_dir):
with open('%s/tasks' % log_dir) as f:
command = next(f).strip()
targets = [target.strip() for target in f]
for target in targets:
out = open_exclusive('%s/%s.out' % (log_dir, sanitize_path(target)))
if not out: continue
cmd_with_target = 'export x=%s; %s' % (target, command)
out.write('%s\n%s\n' % (cmd_with_target, '-'*80))
out.flush()
start_time = datetime.datetime.now()
shell(cmd_with_target, stdout=out, stderr=out)
end_time = datetime.datetime.now()
out.write('%s\nJOB FINISHED. ELAPSED TIME WAS %s.\n' %
('-'*80, end_time - start_time))
out.close()
def cghub_download(samples):
for sample in samples:
# Don't redownload files that are already present.
existing = {}
for root, dirnames, filenames in os.walk('.'):
for f in filenames:
path = os.path.join(root, f)
existing[f] = os.stat(path).st_size
filename = sample.files[0]
filesize = sample.filesizes[0]
if filename in existing and existing[filename] == filesize:
info('%s has already been downloaded...' % filename)
continue
info('Downloading %s...' % filename)
shell('gtdownload -v -d %s -c ~/tools/genetorrent*/cghub_2016.key' %
sample.analysis_data_uri)
def sam_unaligned_reads(bam_path):
# The "samtools bam2fq" command does not output supplementary or
# secondary alignments. Each read has max 1 primary alignment.
options = '-n' if has_mate_suffixes(bam_path) else ''
if has_base_qualities(bam_path):
shell('samtools view -u -f 0x4 -F 0x900 %s | samtools bam2fq %s -' %
(bam_path, options))
else:
bam2fq = shell_stdout(
'samtools view -u -f 0x4 -F 0x900 %s | samtools bam2fq %s -' %
(bam_path, options))
for line in bam2fq:
if line[0] != '@': error('Invalid bam2fq output.')
sys.stdout.write('>')
sys.stdout.write(line[1:])
sys.stdout.write(next(bam2fq))
# Skip per-base qualities. They can start with '@'.
next(bam2fq)
next(bam2fq)
def fasta_from_sra(sra_path):
shell('~/tools/sratoolkit*/fastq-dump --split-3 --gzip %s' % sra_path)
def visualize_splicing(genes, fastq_prefix, out_prefix):
genome_path = '/data/csb/organisms/homo_sapiens/hg19_flat'
bed_path = '/data/csb/organisms/homo_sapiens/ensembl_68/exons.bed'
genes = genes.replace(' ', '').split(',')
min_anchor = 15
read_len = 90
trim = read_len - min_anchor
chromosomes = read_flat_seq('/data/csb/organisms/homo_sapiens/hg19_flat')
donors = []
acceptors = []
exons = []
for line in zopen(bed_path):
cols = line[:-1].split('\t')
if cols[3] in genes:
chr = cols[0] if cols[0].startswith('chr') else 'chr'+cols[0]
chr_seq = chromosomes[chr]
pos = (int(cols[1])+1, int(cols[2]))
if cols[5] == '+':
acceptors.append((chr, '+', pos[0],
chr_seq[pos[0]-1:pos[0]-1+trim]))
donors.append((chr, '+', pos[1], chr_seq[pos[1]-trim:pos[1]]))
elif cols[5] == '-':
acceptors.append((chr, '-', pos[1],
revcomplement(chr_seq[pos[1]-trim:pos[1]])))
donors.append((chr, '-', pos[0],
revcomplement(chr_seq[pos[0]-1:pos[0]-1+trim])))
exons.append(pos)
# Remove duplicate acceptors and donors.
acceptors = list(set(acceptors))
donors = list(set(donors))
exons = list(set(exons))
# Calculate the contiguous genomic sequence
chr = acceptors[0][0]
if any(a[0] != chr for a in acceptors):
error('Genes must be in the same chromosome!')
genome_window = (min(a[2] for a in acceptors)-2000,
max(a[2] for a in acceptors)+2000)
#contig = chromosomes[chr][genome_window[0]:genome_window[1]]
# Calculate junction sequences
class Junction(object):
def __init__(self, name, seq):
self.name = name
self.sequence = seq
self.reads = 0
self.ratio = 0
junctions = defaultdict(list) # Group junctions by donor
for left in donors:
for right in acceptors:
name = '%d[%s]_%d[%s]' % (left[2], left[1], right[2], right[1])
junctions[left].append(Junction(name, left[3] + right[3]))
print('Generated %d junctions.' % (len(donors) * len(acceptors)))
# Build Bowtie index
index_fasta_path = '%s_ref.fa' % out_prefix
index = open(index_fasta_path, 'w')
#index.write('>contig\n%s\n' % contig)
for donor in junctions:
for junc in junctions[donor]:
index.write('>%s\n%s\n' % (junc.name, junc.sequence))
index.close()
shell('/data/csb/tools/bowtie-0.12.9/bowtie-build -q %s %s_index' %
(index_fasta_path, out_prefix))
# Align reads against junctions and tally junction read counts.
shell('bowtie -v1 -B1 -p8 %s_index <(gunzip -c %s_1.fq.gz %s_2.fq.gz) '
'> %s.bowtie' % (out_prefix, fastq_prefix, fastq_prefix, out_prefix))
junction_by_name = {}
for donor in junctions:
for j in junctions[donor]: junction_by_name[j.name] = j
for line in open('%s.bowtie' % out_prefix):
cols = line[:-1].split('\t')
if not '_' in cols[2]: continue
junction_by_name[cols[2]].reads += 1
# Calculate junction power relative to all outgoing links from donor
for donor in junctions:
total = sum(j.reads for j in junctions[donor])
if total <= 0: continue
for j in junctions[donor]:
j.ratio = float(j.reads) / total
if j.reads > 0:
print('%s: %.1f%% (%d)' % (j.name, j.ratio*100, j.reads))
# Check which exons actually participate in the mature transcripts
active_edges = []
for donor in junctions:
for j in junctions[donor]:
if j.ratio < 0.05: continue
active_edges += [int(x[:-3]) for x in j.name.split('_')]
exons = [[ex[0], ex[1], False] for ex in exons]
ties = []
for edge in set(active_edges):
matches = [ex for ex in exons if edge in ex]
if len(matches) == 1: matches[0][2] = True # Unique match, mark active
if len(matches) > 1: ties.append(matches)
for tie in ties:
if not any(ex[2] for ex in tie):
for ex in tie: ex[2] = True # If still tied, mark all tied active
# Print exon map
from svgfig import Rect, Frame, Poly
rects = [Rect(ex[0], 1, ex[1], 2, stroke='none',
fill='whitesmoke', stroke_linejoin='miter')
for ex in exons if not ex[2]]
rects += [Rect(ex[0], 1, ex[1], 2, stroke='none',
fill='black', stroke_linejoin='miter') for ex in exons if ex[2]]
lines = []
for donor in junctions:
for j in junctions[donor]:
start, end = [int(x[:-3]) for x in j.name.split('_')]
lines.append(Poly([(start,2), ((start+end)/2,3), (end,2)],
stroke_opacity=j.ratio))
Frame(genome_window[0], genome_window[1], 0, 10, *(rects+lines),
width=500).SVG().save('%s.svg' % out_prefix)
shell('rm %s_index.* %s_ref.fa' % (out_prefix, out_prefix))
def detect_specific(bam_path, donors_path, acceptors_path, genome_path,
out_prefix, all_reads):
read_len = sam.read_length(bam_path)
info('Using read length %d bp...' % read_len)
flank_len = read_len - 10
chromosomes = read_fasta(genome_path)
donor_exons = regions_from_bed(donors_path)
donors = []
for ex in donor_exons:
chr = ex[0] if ex[0].startswith('chr') else 'chr' + ex[0]
chr_seq = chromosomes[chr]
if ex[1] == '+':
donors.append((chr, '+', ex[3], chr_seq[ex[3] - flank_len:ex[3]]))
elif ex[1] == '-':
donors.append(
(chr, '-', ex[2],
revcomplement(chr_seq[ex[2] - 1:ex[2] - 1 + flank_len])))
acceptor_exons = regions_from_bed(acceptors_path)
acceptors = []
for ex in acceptor_exons:
chr = ex[0] if ex[0].startswith('chr') else 'chr' + ex[0]
chr_seq = chromosomes[chr]
if ex[1] == '+':
acceptors.append(
(chr, '+', ex[2], chr_seq[ex[2] - 1:ex[2] - 1 + flank_len]))
elif ex[1] == '-':
acceptors.append((chr, '-', ex[3],
revcomplement(chr_seq[ex[3] - flank_len:ex[3]])))
del chromosomes # Release 3 GB of memory
gc.collect()
# Remove duplicate acceptors and donors.
acceptors = list(set(acceptors))
donors = list(set(donors))
# Calculate junction sequences
junctions = {}
for left in donors:
for right in acceptors:
name = '%s:%s:%d_%s:%s:%d' % (left[:3] + right[:3])
junctions[name] = Object(sequence=left[3] + right[3], reads=[])
info('Generated %d junctions.' % len(junctions))
# Build Bowtie index
info('Constructing junction FASTA file...')
index_fasta_path = out_prefix + '_ref.fa'
index = open(index_fasta_path, 'w')
for name, junction in junctions.iteritems():
index.write('>%s\n%s\n' % (name, junction.sequence))
index.close()
info('Constructing Bowtie index...')
shell('bowtie-build -q %s %s_index' % (index_fasta_path, out_prefix))
# Align reads against junctions and tally junction read counts.
if all_reads:
info('Aligning all reads against index...')
reads_command = 'sam reads %s' % bam_path
else:
info('Aligning unaligned reads against index...')
reads_command = 'sam unaligned reads %s' % bam_path
for line in shell_stdout('bowtie -f -v1 -B1 %s_index <(%s)' %
(out_prefix, reads_command)):
cols = line.rstrip().split('\t')
junctions[cols[2]].reads.append(cols[4])
shell('rm %s_index.* %s_ref.fa' % (out_prefix, out_prefix))
out_file = open(out_prefix + '.tsv', 'w')
out_file.write('5\' breakpoint\t3\' breakpoint\tNum reads\tSequences\n')
for name, j in junctions.iteritems():
if not j.reads: continue
flanks = name.split('_')
out_file.write('%s\t%s\t%d\t' % (flanks[0], flanks[1], len(j.reads)))
#out_file.write(';'.join(j.reads))
out_file.write('\n')
out_file.close()
def detect_rearrangements(sam_path,
genome_path,
out_prefix,
anchor_len,
min_mapq,
orientation,
max_frag_len,
discard_duplicates='both-ends'):
if not os.path.exists(sam_path):
error('File %s does not exist.' % sam_path)
if not discard_duplicates in ('no', 'both-ends', 'one-end'):
error('Invalid duplicate discard method: %s' % discard_duplicates)
detect_discordant_pairs(sam_path,
out_prefix,
max_frag_len=max_frag_len,
min_mapq=min_mapq,
orientation=orientation)
# Execute split read analysis if the user has specified an anchor length.
if anchor_len > 0:
detect_discordant_reads(sam_path, genome_path, out_prefix, anchor_len)
info('Sorting discordant pairs by chromosomal position...')
sort_inputs = '<(gunzip -c %s.discordant_pairs.tsv.gz)' % out_prefix
if anchor_len > 0:
sort_inputs += ' <(gunzip -c %s.discordant_reads.tsv.gz)' % out_prefix
sort_tmp_dir = os.path.dirname(out_prefix)
if not sort_tmp_dir: sort_tmp_dir = './'
shell('sort -k1,1 -k3,3n -T %s %s | gzip -c > %s.sorted_pairs.tsv.gz' %
(sort_tmp_dir, sort_inputs, out_prefix))
def report_rearrangement(out, r):
if discard_duplicates == 'both-ends':
discard_duplicates_both_ends(r)
elif discard_duplicates == 'one-end':
discard_duplicates_one_end(r)
if len(r.reads) < 2: return 0
out.write('%s\t%s\t%d\t\t\t%s\t%s\t%d\t\t\t%d\t%d\t%s\n' %
(r.chr, r.strand, r.pos, r.mchr, r.mstrand, r.mpos,
sum([read[2] == None for read in r.reads]),
sum([read[2] != None for read in r.reads]), ';'.join(
[read[2] for read in r.reads if read[2] != None])))
return 1
info('Identifying rearrangements based on clusters of discordant reads...')
out = open('%s.sv' % out_prefix, 'w')
out.write(sv_file_header + '\n')
N = 0
rearrangements = []
for line in zopen('%s.sorted_pairs.tsv.gz' % out_prefix):
al = line[:-1].split('\t')
chr = al[0]
strand = al[1]
pos = int(al[2])
mchr = al[3]
mstrand = al[4]
mpos = int(al[5])
seq = None if al[6] == '-' else al[6]
# Rearrangements that are too far need not be considered in the future
reachable = []
for r in rearrangements:
if pos - r.pos > max_frag_len:
N += report_rearrangement(out, r)
else:
reachable.append(r)
rearrangements = reachable
# Check if we already have a rearrangement that matches the new pair.
# We don't check the distance for the first mate because we already
# know from above the rearrangements near it.
matches = [
r for r in rearrangements
if abs(mpos - r.mpos) <= max_frag_len and chr == r.chr
and mchr == r.mchr and strand == r.strand and mstrand == r.mstrand
]
read = (pos, mpos, seq)
if matches:
for match in matches:
match.reads.append(read)
else:
# No suitable rearrangements, create a new one.
rearrangements.append(
Rearrangement(chr, strand, pos, mchr, mstrand, mpos, read))
for r in rearrangements:
N += report_rearrangement(out, r)
info('Found %d rearrangements with at least 2 reads of evidence.' % N)
