def select_transcripts(options):
# load all transcripts
transcript_models_dict = load_data.load_gtf(options.gtf_file)
transcript_models = transcript_models_dict.values()
T = len(transcript_models)
# get translation level in all transcripts
ribo_track = load_data.RiboSeq(options.riboseq_file)
transcript_translation_rate = [
c / float(t.mask.sum()) for c, t in zip(
ribo_track.get_total_counts(transcript_models), transcript_models)
]
# select top transcripts
transcripts = []
transcript_bounds = dict()
order = np.argsort(transcript_translation_rate)[::-1]
for index in order:
transcript = transcript_models[index]
# check if all exons have at least 5 footprints
exon_counts = ribo_track.get_exon_total_counts([transcript])[0]
if np.any(exon_counts < 5):
continue
# check if transcript overlaps any previous transcript
# filter out strict overlaps
overlap = False
try:
for bound in transcript_bounds[transcript.chromosome]:
if not (transcript.stop < bound[0]
or transcript.start > bound[1]):
overlap = True
break
except KeyError:
pass
if overlap:
continue
transcripts.append(transcript)
try:
transcript_bounds[transcript.chromosome].append(
[transcript.start, transcript.stop])
except KeyError:
transcript_bounds[transcript.chromosome] = [[
transcript.start, transcript.stop
]]
# select fixed number of transcripts for learning
if len(transcripts) >= options.batch:
break
return transcripts
def select_transcripts(options):
# load all transcripts
transcript_models_dict = load_data.load_gtf(options.gtf_file)
transcript_models = transcript_models_dict.values()
T = len(transcript_models)
# get translation level in all transcripts
ribo_track = load_data.RiboSeq(options.riboseq_file)
transcript_translation_rate = [c/float(t.mask.sum()) for c,t in
zip(ribo_track.get_total_counts(transcript_models), transcript_models)]
# select top transcripts
transcripts = []
transcript_bounds = dict()
order = np.argsort(transcript_translation_rate)[::-1]
for index in order:
transcript = transcript_models[index]
# check if all exons have at least 5 footprints
exon_counts = ribo_track.get_exon_total_counts([transcript])[0]
if np.any(exon_counts<5):
continue
# check if transcript overlaps any previous transcript
# filter out strict overlaps
overlap = False
try:
for bound in transcript_bounds[transcript.chromosome]:
if not (transcript.stop<bound[0] or transcript.start>bound[1]):
overlap = True
break
except KeyError:
pass
if overlap:
continue
transcripts.append(transcript)
try:
transcript_bounds[transcript.chromosome].append([transcript.start, transcript.stop])
except KeyError:
transcript_bounds[transcript.chromosome] = [[transcript.start, transcript.stop]]
# select fixed number of transcripts for learning
if len(transcripts)>=options.batch:
break
return transcripts
def infer(options):
# load the model
handle = open(options.model_file, 'r')
transition = cPickle.load(handle)
emission = cPickle.load(handle)
handle.close()
# load transcripts
transcript_models = load_data.load_gtf(options.gtf_file)
transcript_names = transcript_models.keys()
N = len(transcript_names)
n = int(np.ceil(N / 1000))
# load data tracks
genome_track = load_data.Genome(options.fasta_file,
options.mappability_file)
ribo_track = load_data.RiboSeq(options.riboseq_file)
if options.rnaseq_file is not None:
rnaseq_track = load_data.RnaSeq(options.rnaseq_file)
# open output file handle
# file in bed12 format
handle = open(options.output_file, 'w')
towrite = [
"chromosome", "start", "stop", "transcript_id", "posterior", "strand",
"cdstart", "cdstop", "protein_seq", "num_exons", "exon_sizes",
"exon_starts"
]
handle.write(" ".join(map(str, towrite)) + '\n')
for n in xrange(N):
tnames = transcript_names[n * 1000:(n + 1) * 1000]
alltranscripts = [transcript_models[name] for name in tnames]
# run inference on both strands independently
# focus on positive strand
for t in alltranscripts:
if t.strand == '-':
t.mask = t.mask[::-1]
t.strand = '+'
# check if all exons have at least 5 footprints
exon_counts = ribo_track.get_exon_total_counts(alltranscripts)
transcripts = [
t for t, e in zip(alltranscripts, exon_counts) if np.all(e >= 5)
]
T = len(transcripts)
if T > 0:
# load sequence of transcripts and transform sequence data
codon_flags = []
rna_sequences = genome_track.get_sequence(transcripts)
for rna_sequence in rna_sequences:
sequence = seq.RnaSequence(rna_sequence)
codon_flags.append(sequence.mark_codons())
# load footprint count data in transcripts
footprint_counts = ribo_track.get_counts(transcripts)
# load transcript-level rnaseq RPKM
if options.rnaseq_file is None:
rna_counts = np.ones((T, ), dtype='float')
else:
rna_counts = rnaseq_track.get_total_counts(transcripts)
# load mappability of transcripts; transform mappability to missingness
if options.mappability_file is not None:
rna_mappability = genome_track.get_mappability(transcripts)
else:
rna_mappability = [
np.ones(c.shape, dtype='bool') for c in footprint_counts
]
# run the learning algorithm
states, frames = ribohmm.infer_coding_sequence(footprint_counts, codon_flags, \
rna_counts, rna_mappability, transition, emission)
# write results
ig = [write_inferred_cds(handle, transcript, state, frame, rna_sequence) \
for transcript,state,frame,rna_sequence in zip(transcripts,states,frames,rna_sequences)]
# focus on negative strand
for t in alltranscripts:
t.mask = t.mask[::-1]
t.strand = '-'
# check if all exons have at least 5 footprints
exon_counts = ribo_track.get_exon_total_counts(alltranscripts)
transcripts = [
t for t, e in zip(alltranscripts, exon_counts) if np.all(e >= 5)
]
T = len(transcripts)
if T > 0:
# load sequence of transcripts and transform sequence data
codon_flags = []
rna_sequences = genome_track.get_sequence(transcripts)
for rna_sequence in rna_sequences:
sequence = seq.RnaSequence(rna_sequence)
codon_flags.append(sequence.mark_codons())
# load footprint count data in transcripts
footprint_counts = ribo_track.get_counts(transcripts)
# load transcript-level rnaseq RPKM
if options.rnaseq_file is None:
rna_counts = np.ones((T, ), dtype='float')
else:
rna_counts = rnaseq_track.get_total_counts(transcripts)
# load mappability of transcripts; transform mappability to missingness
if options.mappability_file is not None:
rna_mappability = genome_track.get_mappability(transcripts)
else:
rna_mappability = [
np.ones(c.shape, dtype='bool') for c in footprint_counts
]
# run the learning algorithm
states, frames = ribohmm.infer_coding_sequence(footprint_counts, codon_flags, \
rna_counts, rna_mappability, transition, emission)
# write results
ig = [write_inferred_cds(handle, transcript, state, frame, rna_sequence) \
for transcript,state,frame,rna_sequence in zip(transcripts,states,frames,rna_sequences)]
handle.close()
ribo_track.close()
if options.rnaseq_file is not None:
rnaseq_track.close()
genome_track.close()
options.output_fastq_prefix = options.gtf_file + '_%d.fq.gz' % options.footprint_length
else:
options.output_fastq_prefix = options.output_fastq_prefix + '_%d.fq.gz' % options.footprint_length
return options
if __name__ == "__main__":
options = parse_args()
qual = ''.join(['~' for r in xrange(options.footprint_length)])
seq_handle = pysam.FastaFile(options.fasta_file)
# load transcripts
transcripts = load_data.load_gtf(options.gtf_file)
tnames = transcripts.keys()
fastq_handle = gzip.open(options.output_fastq_prefix, 'wb')
for num, tname in enumerate(tnames):
transcript = transcripts[tname]
# get transcript DNA sequence
sequence = seq_handle.fetch(transcript.chromosome, transcript.start,
transcript.stop).upper()
# get forward strand reads
if transcript.strand == "-":
transcript.mask = transcript.mask[::-1]
transcript.strand = "+"
if options.output_fastq_prefix is None:
options.output_fastq_prefix = options.gtf_file+'_%d.fq.gz'%options.footprint_length
else:
options.output_fastq_prefix = options.output_fastq_prefix+'_%d.fq.gz'%options.footprint_length
return options
if __name__=="__main__":
options = parse_args()
qual = ''.join(['~' for r in xrange(options.footprint_length)])
seq_handle = pysam.FastaFile(options.fasta_file)
# load transcripts
transcripts = load_data.load_gtf(options.gtf_file)
tnames = transcripts.keys()
fastq_handle = gzip.open(options.output_fastq_prefix, 'wb')
for num,tname in enumerate(tnames):
transcript = transcripts[tname]
# get transcript DNA sequence
sequence = seq_handle.fetch(transcript.chromosome, transcript.start, transcript.stop).upper()
# get forward strand reads
if transcript.strand=="-":
transcript.mask = transcript.mask[::-1]
transcript.strand = "+"
def infer(options):
# load the model
handle = open(options.model_file, 'r')
transition = cPickle.load(handle)
emission = cPickle.load(handle)
handle.close()
# load transcripts
transcript_models = load_data.load_gtf(options.gtf_file)
transcript_names = transcript_models.keys()
N = len(transcript_names)
n = int(np.ceil(N/1000))
# load data tracks
genome_track = load_data.Genome(options.fasta_file, options.mappability_file)
ribo_track = load_data.RiboSeq(options.riboseq_file)
if options.rnaseq_file is not None:
rnaseq_track = load_data.RnaSeq(options.rnaseq_file)
# open output file handle
# file in bed12 format
handle = open(options.output_file,'w')
towrite = ["chromosome", "start", "stop", "transcript_id",
"posterior", "strand", "cdstart", "cdstop",
"protein_seq", "num_exons", "exon_sizes", "exon_starts"]
handle.write(" ".join(map(str,towrite))+'\n')
for n in xrange(N):
tnames = transcript_names[n*1000:(n+1)*1000]
alltranscripts = [transcript_models[name] for name in tnames]
# run inference on both strands independently
# focus on positive strand
for t in alltranscripts:
if t.strand=='-':
t.mask = t.mask[::-1]
t.strand = '+'
# check if all exons have at least 5 footprints
exon_counts = ribo_track.get_exon_total_counts(alltranscripts)
transcripts = [t for t,e in zip(alltranscripts,exon_counts) if np.all(e>=5)]
T = len(transcripts)
if T>0:
# load sequence of transcripts and transform sequence data
codon_flags = []
rna_sequences = genome_track.get_sequence(transcripts)
for rna_sequence in rna_sequences:
sequence = seq.RnaSequence(rna_sequence)
codon_flags.append(sequence.mark_codons())
# load footprint count data in transcripts
footprint_counts = ribo_track.get_counts(transcripts)
# load transcript-level rnaseq RPKM
if options.rnaseq_file is None:
rna_counts = np.ones((T,), dtype='float')
else:
rna_counts = rnaseq_track.get_total_counts(transcripts)
# load mappability of transcripts; transform mappability to missingness
if options.mappability_file is not None:
rna_mappability = genome_track.get_mappability(transcripts)
else:
rna_mappability = [np.ones(c.shape,dtype='bool') for c in footprint_counts]
# run the learning algorithm
states, frames = ribohmm.infer_coding_sequence(footprint_counts, codon_flags, \
rna_counts, rna_mappability, transition, emission)
# write results
ig = [write_inferred_cds(handle, transcript, state, frame, rna_sequence) \
for transcript,state,frame,rna_sequence in zip(transcripts,states,frames,rna_sequences)]
# focus on negative strand
for t in alltranscripts:
t.mask = t.mask[::-1]
t.strand = '-'
# check if all exons have at least 5 footprints
exon_counts = ribo_track.get_exon_total_counts(alltranscripts)
transcripts = [t for t,e in zip(alltranscripts,exon_counts) if np.all(e>=5)]
T = len(transcripts)
if T>0:
# load sequence of transcripts and transform sequence data
codon_flags = []
rna_sequences = genome_track.get_sequence(transcripts)
for rna_sequence in rna_sequences:
sequence = seq.RnaSequence(rna_sequence)
codon_flags.append(sequence.mark_codons())
# load footprint count data in transcripts
footprint_counts = ribo_track.get_counts(transcripts)
# load transcript-level rnaseq RPKM
if options.rnaseq_file is None:
rna_counts = np.ones((T,), dtype='float')
else:
rna_counts = rnaseq_track.get_total_counts(transcripts)
# load mappability of transcripts; transform mappability to missingness
if options.mappability_file is not None:
rna_mappability = genome_track.get_mappability(transcripts)
else:
rna_mappability = [np.ones(c.shape,dtype='bool') for c in footprint_counts]
# run the learning algorithm
states, frames = ribohmm.infer_coding_sequence(footprint_counts, codon_flags, \
rna_counts, rna_mappability, transition, emission)
# write results
ig = [write_inferred_cds(handle, transcript, state, frame, rna_sequence) \
for transcript,state,frame,rna_sequence in zip(transcripts,states,frames,rna_sequences)]
handle.close()
ribo_track.close()
if options.rnaseq_file is not None:
rnaseq_track.close()
genome_track.close()