def to_gtf(self):
strand_str = Strand.to_gtf(self.strand)
f = GTF.Feature()
f.seqid = self.chrom
f.source = 'taco'
f.feature = 'transcript'
f.start = self.start
f.end = self.end
f.score = 0.0
f.strand = strand_str
f.phase = '.'
f.attrs = {
GTF.Attr.TRANSCRIPT_ID: self._id,
GTF.Attr.SAMPLE_ID: self.sample_id,
GTF.Attr.EXPR: str(self.expr),
GTF.Attr.REF: str(int(self.is_ref))
}
yield f
for e in self.exons:
f = GTF.Feature()
f.seqid = self.chrom
f.source = 'taco'
f.feature = 'exon'
f.start = e.start
f.end = e.end
f.score = 0.0
f.strand = strand_str
f.phase = '.'
f.attrs = {GTF.Attr.TRANSCRIPT_ID: self._id}
yield f
def get_change_point_gtf(self, cp):
graph_id = ('G_%s_%d_%d_%s' %
(self.chrom, self.start, self.end,
Strand.to_gtf(self.strand)))
features = []
f = GTF.Feature()
f.seqid = self.chrom
f.source = 'taco'
f.feature = 'changept'
f.start = cp.pos
f.end = cp.pos + 1
f.score = 0
f.strand = Strand.to_gtf(self.strand)
f.phase = '.'
f.attrs = {'graph_id': graph_id,
'sign': str(cp.sign),
'pvalue': str(cp.pvalue),
'foldchange': str(cp.foldchange)}
features.append(f)
f = GTF.Feature()
f.seqid = self.chrom
f.source = 'taco'
f.feature = 'changeinterval'
f.start = cp.start
f.end = cp.end
f.score = 0
f.strand = Strand.to_gtf(self.strand)
f.phase = '.'
f.attrs = {'graph_id': graph_id,
'sign': str(cp.sign),
'pvalue': str(cp.pvalue),
'foldchange': str(cp.foldchange)}
features.append(f)
return features
def get_node_gtf(self):
graph_id = (
'G_%s_%d_%d_%s' %
(self.chrom, self.start, self.end, Strand.to_gtf(self.strand)))
# iterate through locus and return change point data
for n_id in self.G:
n = self.get_node_interval(n_id)
expr_data = self.get_expr_data(*n)
ref_starts = _array_subset(self.ref_start_sites, *n)
ref_stops = _array_subset(self.ref_stop_sites, *n)
# return gtf feature for each node
f = GTF.Feature()
f.seqid = self.chrom
f.source = 'taco'
f.feature = 'node'
f.start = n[0]
f.end = n[1]
f.score = 0
f.strand = Strand.to_gtf(self.strand)
f.phase = '.'
f.attrs = {
'graph_id': graph_id,
'expr_min': str(expr_data.min()),
'expr_max': str(expr_data.max()),
'expr_mean': str(expr_data.mean()),
'ref_starts': ','.join(map(str, ref_starts)),
'ref_stops': ','.join(map(str, ref_stops))
}
yield f
def get_gtf_features(chrom, strand, exons, locus_id, gene_id, tss_id,
transcript_id, expr, rel_frac, abs_frac):
tx_start = exons[0].start
tx_end = exons[-1].end
strand_str = Strand.to_gtf(strand)
attr_dict = {
'locus_id': locus_id,
'gene_id': gene_id,
'tss_id': tss_id,
'transcript_id': transcript_id
}
f = GTF.Feature()
f.seqid = chrom
f.source = 'taco'
f.feature = 'transcript'
f.start = tx_start
f.end = tx_end
f.score = int(round(1000.0 * rel_frac))
f.strand = strand_str
f.phase = '.'
f.attrs = {
'expr': '%.3f' % expr,
'rel_frac': '%.5f' % rel_frac,
'abs_frac': '%.5f' % abs_frac
}
f.attrs.update(attr_dict)
yield f
for e in exons:
f = GTF.Feature()
f.seqid = chrom
f.source = 'taco'
f.feature = 'exon'
f.start = e.start
f.end = e.end
f.score = int(round(1000.0 * rel_frac))
f.strand = strand_str
f.phase = '.'
f.attrs = {}
f.attrs.update(attr_dict)
yield f
def _make_transcript_feature(exon_features):
f = GTF.Feature()
f.seqid = exon_features[0].seqid
f.source = exon_features[0].source
f.feature = 'transcript'
f.start = exon_features[0].start
f.end = exon_features[-1].end
f.score = exon_features[0].score
f.strand = exon_features[0].strand
f.phase = '.'
f.attrs = exon_features[0].attrs.copy()
if 'exon_number' in f.attrs:
del f.attrs['exon_number']
return f
def assemble_isoforms(sgraph, config):
# create a path graph from the splice graph
K, k = create_optimal_path_graph(
sgraph,
kmax=config.path_graph_kmax,
loss_threshold=config.path_graph_loss_threshold,
stats_fh=config.path_graph_stats_fh)
if K is None:
return []
if len(K) == 0:
return []
# report lost nodes
if config.assembly_loss_gtf_fh is not None:
graph_id = ('L_%s:%d-%d[%s]' % (sgraph.chrom, sgraph.start, sgraph.end,
Strand.to_gtf(sgraph.strand)))
for n_id in get_lost_nodes(sgraph, K):
n = sgraph.get_node_interval(n_id)
expr_data = sgraph.get_node_expr_data(n_id)
# return gtf feature for each node
f = GTF.Feature()
f.seqid = sgraph.chrom
f.source = 'taco'
f.feature = 'lost_node'
f.start = n[0]
f.end = n[1]
f.score = 0.0
f.strand = Strand.to_gtf(sgraph.strand)
f.phase = '.'
f.attrs = {'graph_id': graph_id, 'expr': str(expr_data.mean())}
print >> config.assembly_loss_gtf_fh, str(f)
# smooth kmer graph
smooth_graph(K)
source_node = K.graph['source']
source_expr = K.node[source_node][KMER_EXPR]
logging.debug('%s:%d-%d[%s] finding paths in k=%d graph '
'(%d nodes) source_expr=%f' %
(sgraph.chrom, sgraph.start, sgraph.end,
Strand.to_gtf(sgraph.strand), k, len(K), source_expr))
id_kmer_map = K.graph['id_kmer_map']
paths = []
for kmer_path, expr in find_paths(K, KMER_EXPR, config.path_frac,
config.max_paths):
path = reconstruct_path(kmer_path, id_kmer_map, sgraph)
logging.debug("\texpr=%f length=%d" % (expr, len(path)))
paths.append((path, expr))
# build gene clusters
clusters, filtered = Cluster.build(paths, min_frac=config.isoform_frac)
logging.debug('\tclusters: %d filtered: %d' %
(len(clusters), len(filtered)))
gene_isoforms = []
for cluster in clusters:
isoforms = []
for path, expr, rel_frac, abs_frac in cluster.iterpaths():
isoforms.append(
Isoform(path=path,
expr=expr,
rel_frac=rel_frac,
abs_frac=abs_frac))
# apply max isoforms limit (per cluster)
if config.max_isoforms > 0:
isoforms = isoforms[:config.max_isoforms]
gene_isoforms.append(isoforms)
return gene_isoforms
