def get_ids_passing_filter(gff_index_dir,
bam_filename,
output_dir):
"""
Apply filter to events using bedtools and return
only the events that meet the filter.
"""
min_reads = 20
settings = Settings.get()
min_event_reads = Settings.get_min_event_reads()
# Check that this was indexed with a version that outputs
# genes.gff file
genes_gff_fname = os.path.join(gff_index_dir,
"genes.gff")
if not os.path.isfile(genes_gff_fname):
print "WARNING: Could not find \'genes.gff\' in %s - " \
"skipping prefilter stage. Please reindex your " \
"GFF file with the latest version to enable " \
"prefiltering." %(gff_index_dir)
return None
print "Prefiltering reads..."
coverage_fname = exon_utils.get_bam_gff_coverage(bam_filename,
genes_gff_fname,
output_dir)
ids_passing_filter = []
with open(coverage_fname) as coverage_in:
for line in coverage_in:
# Skip comments
if line.startswith("#"):
continue
fields = line.strip().split("\t")
# Get the counts field and the event ID
# if it passes the filter
counts = int(fields[9])
if counts < min_event_reads:
continue
attribs = gff_utils.parse_gff_attribs(fields[8])
if "ID" not in attribs:
print "WARNING: No ID= found for line:\n%s\nSkipping..." \
%(line)
continue
event_id = attribs["ID"]
ids_passing_filter.append(event_id)
return ids_passing_filter
def get_ids_passing_filter(gff_index_dir,
bam_filename,
output_dir):
"""
Apply filter to events using bedtools and return
only the events that meet the filter.
"""
min_reads = 20
settings = Settings.get()
min_event_reads = Settings.get_min_event_reads()
# Check that this was indexed with a version that outputs
# genes.gff file
genes_gff_fname = os.path.join(gff_index_dir,
"genes.gff")
if not os.path.isfile(genes_gff_fname):
print "WARNING: Could not find \'genes.gff\' in %s - " \
"skipping prefilter stage. Please reindex your " \
"GFF file with the latest version to enable " \
"prefiltering." %(gff_index_dir)
return None
print "Prefiltering reads..."
coverage_fname = exon_utils.get_bam_gff_coverage(bam_filename,
genes_gff_fname,
output_dir)
ids_passing_filter = []
with open(coverage_fname) as coverage_in:
for line in coverage_in:
# Skip comments
if line.startswith("#"):
continue
fields = line.strip().split("\t")
# Get the counts field and the event ID
# if it passes the filter
counts = int(fields[9])
if counts < min_event_reads:
continue
attribs = gff_utils.parse_gff_attribs(fields[8])
if "ID" not in attribs:
print "WARNING: No ID= found for line:\n%s\nSkipping..." \
%(line)
continue
event_id = attribs["ID"]
ids_passing_filter.append(event_id)
return ids_passing_filter
def compute_gene_psi(gene_ids, gff_index_filename, bam_filename,
output_dir, read_len, overhang_len,
paired_end=None,
event_type=None,
verbose=True):
"""
Run Psi at the Gene-level (for multi-isoform inference.)
Arguments:
- Set of gene IDs corresponding to gene IDs from the GFF
- Indexed GFF filename describing the genes
- BAM filename with the reads (must be sorted and indexed)
- Output directory
- Optional: Run in paired-end mode. Gives mean and standard deviation
of fragment length distribution.
"""
misc_utils.make_dir(output_dir)
if not os.path.exists(gff_index_filename):
print "Error: No GFF %s" %(gff_index_filename)
return
num_genes = len(gene_ids)
print "Computing Psi for %d genes..." %(num_genes)
print " - " + ", ".join(gene_ids)
print " - GFF filename: %s" %(gff_index_filename)
print " - BAM: %s" %(bam_filename)
print " - Outputting to: %s" %(output_dir)
if paired_end:
print " - Paired-end mode: ", paired_end
settings = Settings.get()
settings_params = Settings.get_sampler_params()
burn_in = settings_params["burn_in"]
lag = settings_params["lag"]
num_iters = settings_params["num_iters"]
num_chains = settings_params["num_chains"]
min_event_reads = Settings.get_min_event_reads()
strand_rule = Settings.get_strand_param()
mean_frag_len = None
frag_variance = None
if paired_end:
mean_frag_len = int(paired_end[0])
frag_variance = power(int(paired_end[1]), 2)
# Load the genes from the GFF
gff_genes = gff_utils.load_indexed_gff_file(gff_index_filename)
# If given a template for the SAM file, use it
template = None
if settings and "sam_template" in settings:
template = settings["sam_template"]
if "filter_reads" not in settings:
filter_reads = True
else:
filter_reads = settings["filter_reads"]
# Load the BAM file upfront
bamfile = sam_utils.load_bam_reads(bam_filename,
template=template)
# Check if we're in compressed mode
compressed_mode = misc_utils.is_compressed_index(gff_index_filename)
for gene_id, gene_info in gff_genes.iteritems():
lookup_id = gene_id
# Skip genes that we were not asked to run on
if lookup_id not in gene_ids:
continue
gene_obj = gene_info['gene_object']
gene_hierarchy = gene_info['hierarchy']
# Sanity check: if the isoforms are all shorter than the read,
# skip the event
if all(map(lambda l: l < read_len, gene_obj.iso_lens)):
print "All isoforms of %s shorter than %d, so skipping" \
%(gene_id, read_len)
continue
# Find the most inclusive transcription start and end sites
# for each gene
tx_start, tx_end = \
gff_utils.get_inclusive_txn_bounds(gene_info['hierarchy'][gene_id])
# Fetch reads aligning to the gene boundaries
gene_reads = \
sam_utils.fetch_bam_reads_in_gene(bamfile,
gene_obj.chrom,
tx_start,
tx_end,
gene_obj)
# Parse reads: checking strandedness and pairing
# reads in case of paired-end data
reads, num_raw_reads = \
sam_utils.sam_parse_reads(gene_reads,
paired_end=paired_end,
strand_rule=strand_rule,
target_strand=gene_obj.strand,
given_read_len=read_len)
# Skip gene if none of the reads align to gene boundaries
if filter_reads:
if num_raw_reads < min_event_reads:
print "Only %d reads in gene, skipping (needed >= %d reads)" \
%(num_raw_reads,
min_event_reads)
continue
else:
print "%d raw reads in event" %(num_raw_reads)
num_isoforms = len(gene_obj.isoforms)
hyperparameters = ones(num_isoforms)
##
## Run the sampler
##
# Create the sampler with the right parameters depending on whether
# this is a paired-end or single-end data set.
if paired_end:
# Sampler parameters for paired-end mode
sampler_params = \
miso.get_paired_end_sampler_params(num_isoforms,
mean_frag_len,
frag_variance,
read_len,
overhang_len=overhang_len)
sampler = miso.MISOSampler(sampler_params,
paired_end=True,
log_dir=output_dir)
else:
# Sampler parameters for single-end mode
sampler_params = miso.get_single_end_sampler_params(num_isoforms,
read_len,
overhang_len)
sampler = miso.MISOSampler(sampler_params,
paired_end=False,
log_dir=output_dir)
# Make directory for chromosome -- if given an event type, put
# the gene in the event type directory
if event_type != None:
chrom_dir = os.path.join(output_dir, event_type, gene_obj.chrom)
else:
chrom_dir = os.path.join(output_dir, gene_obj.chrom)
try:
os.makedirs(chrom_dir)
except OSError:
pass
# Pick .miso output filename based on the pickle filename
miso_basename = os.path.basename(gff_index_filename)
if not miso_basename.endswith(".pickle"):
print "Error: Invalid index file %s" %(gff_index_filename)
sys.exit(1)
miso_basename = miso_basename.replace(".pickle", "")
output_filename = os.path.join(chrom_dir, "%s" %(miso_basename))
sampler.run_sampler(num_iters, reads, gene_obj, hyperparameters,
sampler_params, output_filename,
num_chains=num_chains,
burn_in=burn_in,
lag=lag)
def __init__(self, gff_dir, bam_filename,
output_dir, read_len, overhang_len,
main_logger,
settings_fname=None,
paired_end=None,
use_cluster=False,
chunk_jobs=200,
SGEarray=False,
sge_job_name="misojob",
gene_ids=None,
num_proc=None,
wait_on_jobs=True):
self.main_logger = main_logger
self.threads = {}
self.gff_dir = gff_dir
self.bam_filename = bam_filename
# Check that the BAM filename exists and that it has an index
if not os.path.isfile(self.bam_filename):
self.main_logger.error("BAM file %s not found." %(self.bam_filename))
sys.exit(1)
self.bam_index_fname = "%s.bai" %(self.bam_filename)
if not os.path.isfile(self.bam_index_fname):
self.main_logger.warning("Expected BAM index file %s not found." \
%(self.bam_index_fname))
self.main_logger.warning("Are you sure your BAM file is indexed?")
self.output_dir = output_dir
self.read_len = read_len
# For now setting overhang to 1 always
#self.overhang_len = overhang_len
self.overhang_len = 1
self.settings_fname = settings_fname
self.paired_end = paired_end
self.use_cluster = use_cluster
self.chunk_jobs = chunk_jobs
self.settings = Settings.get()
self.cluster_cmd = Settings.get_cluster_command()
self.sge_job_name = sge_job_name
self.wait_on_jobs = wait_on_jobs
# if chunk_jobs not given (i.e. set to False),
# then set it to arbitrary value
if not self.chunk_jobs:
self.chunk_jobs = 200
self.SGEarray = SGEarray
self.num_processors = Settings.get_num_processors()
if num_proc is not None:
num_proc = int(num_proc)
self.num_processors = num_proc
self.main_logger.info("Using %d processors" %(num_proc))
self.long_thresh = 50
self.batch_logs_dir = \
os.path.join(output_dir, "batch-logs")
self.batch_genes_dir = \
os.path.join(output_dir, "batch-genes")
self.cluster_scripts_dir = \
os.path.join(output_dir, "cluster_scripts")
self.scripts_output_dir = \
os.path.join(output_dir, "scripts_output")
misc_utils.make_dir(self.batch_logs_dir)
misc_utils.make_dir(self.batch_genes_dir)
misc_utils.make_dir(self.cluster_scripts_dir)
misc_utils.make_dir(self.scripts_output_dir)
# First compile a set of genes that should be run on
# and output them to file along with their indexed
# filenames
self.gene_ids_to_gff_index = \
gff_utils.get_gene_ids_to_gff_index(gff_dir)
# If we're given filtered gene IDs, use them
if gene_ids is not None:
self.gene_ids = gene_ids
else:
self.gene_ids = self.gene_ids_to_gff_index.keys()
if len(self.gene_ids) == 0:
self.main_logger.error("No genes to run on. Did you pass me the wrong path " \
"to your index GFF directory? " \
"Or perhaps your indexed GFF directory " \
"is empty?")
sys.exit(1)
self.batch_filenames = self.output_batch_files()
def compute_psi(sample_filenames,
output_dir,
event_type,
read_len,
overhang_len,
use_cluster=False,
chunk_jobs=False,
filter_events=True,
events_info_filename=None,
settings_filename=None):
"""
Compute Psi values for skipped exons. Sample filenames is a mapping from
sample label to sample.
- sample_filenames = [[sample_label1, sample_filename1],
[sample_label2, sample_filename2]]
- output_dir: output directory
- event_type: 'SE', 'RI', etc.
"""
misc_utils.make_dir(output_dir)
output_dir = os.path.join(output_dir, event_type)
output_dir = os.path.abspath(output_dir)
misc_utils.make_dir(output_dir)
print "Computing Psi for events of type %s" % (event_type)
print " - samples used: ", sample_filenames.keys()
for sample_label, sample_filename in sample_filenames.iteritems():
print "Processing sample: label=%s, filename=%s" \
%(sample_label, sample_filename)
results_output_dir = os.path.join(output_dir, sample_label)
misc_utils.make_dir(results_output_dir)
# Load the set of counts and serialize them into JSON
events = \
as_events.load_event_counts(sample_filename,
event_type,
events_info_filename=events_info_filename)
# Filter events
if filter_events:
print "Filtering events..."
events.filter_events(settings=Settings.get())
print "Running on a total of %d events." % (len(events.events))
events_filename = events.output_file(results_output_dir, sample_label)
# Run MISO on them
miso_cmd = "python %s --compute-two-iso-psi %s %s --event-type %s " \
"--read-len %d --overhang-len %d " \
%(os.path.join(miso_path, 'run_miso.py'),
events_filename,
results_output_dir,
event_type,
read_len,
overhang_len)
if use_cluster:
if chunk_jobs:
miso_cmd += ' --use-cluster --chunk-jobs %d' % (chunk_jobs)
else:
miso_cmd += ' --use-cluster'
print "Executing: %s" % (miso_cmd)
if use_cluster:
print " - Using cluster"
os.system(miso_cmd)
def __init__(self, gff_dir, bam_filename,
output_dir, read_len, overhang_len,
settings_fname=None,
paired_end=None,
use_cluster=False,
chunk_jobs=200,
SGEarray=False,
sge_job_name="misojob",
gene_ids=None,
num_proc=None,
wait_on_jobs=True):
self.threads = {}
self.gff_dir = gff_dir
self.bam_filename = bam_filename
# Check that the BAM filename exists and that it has an index
if not os.path.isfile(self.bam_filename):
print "Error: BAM file %s not found." %(self.bam_filename)
sys.exit(1)
self.bam_index_fname = "%s.bai" %(self.bam_filename)
if not os.path.isfile(self.bam_index_fname):
print "WARNING: Expected BAM index file %s not found." \
%(self.bam_index_fname)
print "Are you sure your BAM file is indexed?"
self.output_dir = output_dir
self.read_len = read_len
# For now setting overhang to 1 always
#self.overhang_len = overhang_len
self.overhang_len = 1
self.settings_fname = settings_fname
self.paired_end = paired_end
self.use_cluster = use_cluster
self.chunk_jobs = chunk_jobs
self.settings = Settings.get()
self.cluster_cmd = Settings.get_cluster_command()
self.sge_job_name = sge_job_name
self.wait_on_jobs = wait_on_jobs
# if chunk_jobs not given (i.e. set to False),
# then set it to arbitrary value
if not self.chunk_jobs:
self.chunk_jobs = 200
self.SGEarray = SGEarray
self.num_processors = Settings.get_num_processors()
if num_proc is not None:
num_proc = int(num_proc)
self.num_processors = num_proc
print "Using %d processors" %(num_proc)
self.long_thresh = 50
self.batch_logs_dir = \
os.path.join(output_dir, "batch-logs")
self.batch_genes_dir = \
os.path.join(output_dir, "batch-genes")
self.cluster_scripts_dir = \
os.path.join(output_dir, "cluster_scripts")
self.scripts_output_dir = \
os.path.join(output_dir, "scripts_output")
misc_utils.make_dir(self.batch_logs_dir)
misc_utils.make_dir(self.batch_genes_dir)
misc_utils.make_dir(self.cluster_scripts_dir)
misc_utils.make_dir(self.scripts_output_dir)
# First compile a set of genes that should be run on
# and output them to file along with their indexed
# filenames
self.gene_ids_to_gff_index = \
gff_utils.get_gene_ids_to_gff_index(gff_dir)
# If we're given filtered gene IDs, use them
if gene_ids is not None:
self.gene_ids = gene_ids
else:
self.gene_ids = self.gene_ids_to_gff_index.keys()
if len(self.gene_ids) == 0:
print "Error: No genes to run on. Did you pass me the wrong path " \
"to your index GFF directory? " \
"Or perhaps your indexed GFF directory " \
"is empty?"
sys.exit(1)
self.batch_filenames = self.output_batch_files()
def compute_psi(sample_filenames, output_dir, event_type,
read_len, overhang_len,
use_cluster=False,
chunk_jobs=False,
filter_events=True,
events_info_filename=None,
settings_filename=None):
"""
Compute Psi values for skipped exons. Sample filenames is a mapping from
sample label to sample.
- sample_filenames = [[sample_label1, sample_filename1],
[sample_label2, sample_filename2]]
- output_dir: output directory
- event_type: 'SE', 'RI', etc.
"""
misc_utils.make_dir(output_dir)
output_dir = os.path.join(output_dir, event_type)
output_dir = os.path.abspath(output_dir)
misc_utils.make_dir(output_dir)
print "Computing Psi for events of type %s" %(event_type)
print " - samples used: ", sample_filenames.keys()
for sample_label, sample_filename in sample_filenames.iteritems():
print "Processing sample: label=%s, filename=%s" \
%(sample_label, sample_filename)
results_output_dir = os.path.join(output_dir, sample_label)
misc_utils.make_dir(results_output_dir)
# Load the set of counts and serialize them into JSON
events = \
as_events.load_event_counts(sample_filename,
event_type,
events_info_filename=events_info_filename)
# Filter events
if filter_events:
print "Filtering events..."
events.filter_events(settings=Settings.get())
print "Running on a total of %d events." %(len(events.events))
events_filename = events.output_file(results_output_dir,
sample_label)
# Run MISO on them
miso_cmd = "python %s --compute-two-iso-psi %s %s --event-type %s " \
"--read-len %d --overhang-len %d " \
%(os.path.join(miso_path, 'run_miso.py'),
events_filename,
results_output_dir,
event_type,
read_len,
overhang_len)
if use_cluster:
if chunk_jobs:
miso_cmd += ' --use-cluster --chunk-jobs %d' %(chunk_jobs)
else:
miso_cmd += ' --use-cluster'
print "Executing: %s" %(miso_cmd)
if use_cluster:
print " - Using cluster"
os.system(miso_cmd)
def compute_gene_psi(gene_ids, gff_index_filename, bam_filename,
output_dir, read_len, overhang_len,
paired_end=None,
event_type=None,
verbose=True):
"""
Run Psi at the Gene-level (for multi-isoform inference.)
Arguments:
- Set of gene IDs corresponding to gene IDs from the GFF
- Indexed GFF filename describing the genes
- BAM filename with the reads (must be sorted and indexed)
- Output directory
- Optional: Run in paired-end mode. Gives mean and standard deviation
of fragment length distribution.
"""
misc_utils.make_dir(output_dir)
if not os.path.exists(gff_index_filename):
print "Error: No GFF %s" %(gff_index_filename)
return
num_genes = len(gene_ids)
print "Computing Psi for %d genes..." %(num_genes)
print " - " + ", ".join(gene_ids)
print " - GFF filename: %s" %(gff_index_filename)
print " - BAM: %s" %(bam_filename)
print " - Outputting to: %s" %(output_dir)
if paired_end:
print " - Paired-end mode: ", paired_end
settings = Settings.get()
settings_params = Settings.get_sampler_params()
burn_in = settings_params["burn_in"]
lag = settings_params["lag"]
num_iters = settings_params["num_iters"]
num_chains = settings_params["num_chains"]
min_event_reads = Settings.get_min_event_reads()
strand_rule = Settings.get_strand_param()
mean_frag_len = None
frag_variance = None
if paired_end:
mean_frag_len = int(paired_end[0])
frag_variance = power(int(paired_end[1]), 2)
# Load the genes from the GFF
gff_genes = gff_utils.load_indexed_gff_file(gff_index_filename)
# If given a template for the SAM file, use it
template = None
if settings and "sam_template" in settings:
template = settings["sam_template"]
if "filter_reads" not in settings:
filter_reads = True
else:
filter_reads = settings["filter_reads"]
# Load the BAM file upfront
bamfile = sam_utils.load_bam_reads(bam_filename,
template=template)
# Check if we're in compressed mode
compressed_mode = misc_utils.is_compressed_index(gff_index_filename)
for gene_id, gene_info in gff_genes.iteritems():
lookup_id = gene_id
# Skip genes that we were not asked to run on
if lookup_id not in gene_ids:
continue
gene_obj = gene_info['gene_object']
gene_hierarchy = gene_info['hierarchy']
# Sanity check: if the isoforms are all shorter than the read,
# skip the event
if all(map(lambda l: l < read_len, gene_obj.iso_lens)):
print "All isoforms of %s shorter than %d, so skipping" \
%(gene_id, read_len)
continue
# Find the most inclusive transcription start and end sites
# for each gene
tx_start, tx_end = \
gff_utils.get_inclusive_txn_bounds(gene_info['hierarchy'][gene_id])
# Fetch reads aligning to the gene boundaries
gene_reads = \
sam_utils.fetch_bam_reads_in_gene(bamfile,
gene_obj.chrom,
tx_start,
tx_end,
gene_obj)
# Parse reads: checking strandedness and pairing
# reads in case of paired-end data
reads, num_raw_reads = \
sam_utils.sam_parse_reads(gene_reads,
paired_end=paired_end,
strand_rule=strand_rule,
target_strand=gene_obj.strand)
# Skip gene if none of the reads align to gene boundaries
if filter_reads:
if num_raw_reads < min_event_reads:
print "Only %d reads in gene, skipping (needed >= %d reads)" \
%(num_raw_reads,
min_event_reads)
continue
else:
print "%d raw reads in event" %(num_raw_reads)
num_isoforms = len(gene_obj.isoforms)
hyperparameters = ones(num_isoforms)
##
## Run the sampler
##
# Create the sampler with the right parameters depending on whether
# this is a paired-end or single-end data set.
if paired_end:
# Sampler parameters for paired-end mode
sampler_params = \
miso.get_paired_end_sampler_params(num_isoforms,
mean_frag_len,
frag_variance,
read_len,
overhang_len=overhang_len)
sampler = miso.MISOSampler(sampler_params,
paired_end=True,
log_dir=output_dir)
else:
# Sampler parameters for single-end mode
sampler_params = miso.get_single_end_sampler_params(num_isoforms,
read_len,
overhang_len)
sampler = miso.MISOSampler(sampler_params,
paired_end=False,
log_dir=output_dir)
# Make directory for chromosome -- if given an event type, put
# the gene in the event type directory
if event_type != None:
chrom_dir = os.path.join(output_dir, event_type, gene_obj.chrom)
else:
chrom_dir = os.path.join(output_dir, gene_obj.chrom)
try:
os.makedirs(chrom_dir)
except OSError:
pass
# Pick .miso output filename based on the pickle filename
miso_basename = os.path.basename(gff_index_filename)
if not miso_basename.endswith(".pickle"):
print "Error: Invalid index file %s" %(gff_index_filename)
sys.exit(1)
miso_basename = miso_basename.replace(".pickle", "")
output_filename = os.path.join(chrom_dir, "%s" %(miso_basename))
sampler.run_sampler(num_iters, reads, gene_obj, hyperparameters,
sampler_params, output_filename,
num_chains=num_chains,
burn_in=burn_in,
lag=lag)
