def _get_counts_and_sequence(gtf_iterator, bam, fasta,
seperate_UTRs=False):
'''Called by pentamer_enrichment. This function will return an iterator
that yeilds tuples of profiles accross transcripts or introns and the
sequence for which the profile is determined'''
for transcript in gtf_iterator:
E.debug("Counting transcript %s" % transcript[0].transcript_id)
contig, strand = transcript[0].contig, transcript[0].strand
# exons
exons = GTF.asRanges(transcript, "exon")
sequence = "".join(fasta.getSequence(contig, strand, exon[0], exon[1])
for exon in exons)
exon_counts = count_transcript(transcript, bam)
yield (exon_counts, sequence)
# introns
intron_intervals = GTF.toIntronIntervals(transcript)
intron_counts = count_intervals(bam, intron_intervals, contig, strand)
if intron_counts.sum() == 0:
continue
for intron in intron_intervals:
seq = fasta.getSequence(contig, strand, intron[0], intron[1])
profile = intron_counts.loc[float(intron[0]):float(intron[1])]
profile.index = profile.index - intron[0]
yield (profile, seq)
def _get_counts_and_sequence(gtf_iterator, bam, fasta, seperate_UTRs=False):
'''Called by pentamer_enrichment. This function will return an iterator
that yeilds tuples of profiles accross transcripts or introns and the
sequence for which the profile is determined'''
for transcript in gtf_iterator:
E.debug("Counting transcript %s" % transcript[0].transcript_id)
contig, strand = transcript[0].contig, transcript[0].strand
# exons
exons = GTF.asRanges(transcript, "exon")
sequence = "".join(
fasta.getSequence(contig, strand, exon[0], exon[1])
for exon in exons)
exon_counts = count_transcript(transcript, bam)
yield (exon_counts, sequence)
# introns
intron_intervals = GTF.toIntronIntervals(transcript)
intron_counts = count_intervals(bam, intron_intervals, contig, strand)
if intron_counts.sum() == 0:
continue
for intron in intron_intervals:
seq = fasta.getSequence(contig, strand, intron[0], intron[1])
profile = intron_counts.loc[float(intron[0]):float(intron[1])]
profile.index = profile.index - intron[0]
yield (profile, seq)
def processing_index(interval_iterator, bam, window_size=50):
'''Calculate the processing index for the speicied sample, using the
provided interval_iterator to get the cleavage sites. The iterator
can be GTF or BED, as long as it has end, contig and strand
attributes. The end attribute will be used to define the
cleavage site.
The proccessing index for G genes is defined as:
.. math::
pi = log_2( \frac{\sum_{i=1}^{G} N_i^{PM}}{\sum_{i=1}^{G} N_i^M})
after Baejen et al Mol Cell 5(55):745-757. However, Beaejen et al
normalise this number to the total number of genes, which seems
wrong to me. '''
n_pm = 0
n_m = 0
for site in interval_iterator:
if site.strand == "+":
pos = site.end
elif site.strand == "-":
pos = site.start
else:
raise ValueError(
"processing index not valid for unstranded cleavage points in "
"entry\n" + str(site)+"\n")
upstream_interval = (pos - window_size, pos)
downstream_interval = (pos, pos + window_size)
counts = count_intervals(bam, [upstream_interval, downstream_interval],
site.contig, site.strand)
# We are currently in genome cooridinates, not transcript
if site.strand == "+":
# pandas indexing is inclusive
n_up = counts.iloc[:pos-1].sum()
n_down = counts.iloc[pos:].sum()
elif site.strand == "-":
n_up = counts.iloc[pos:].sum()
n_down = counts.iloc[:pos-1].sum()
n_pm += n_down
n_m += n_up-n_down
pi = np.log2(float(n_pm)/float(max(1, n_m)))
return pi
def _get_counts_and_sequence(gtf_iterator, bam, fasta,
seperate_UTRs=False):
'''Called by pentamer_enrichment. This function will return an iterator
that yeilds tuples of profiles accross transcripts or introns and the
sequence for which the profile is determined'''
for transcript in gtf_iterator:
transcript = [e for e in transcript if hasattr(e, "transcript_id")]
if len(transcript)==0:
continue
contig, strand = transcript[0].contig, transcript[0].strand
# exons
exons = GTF.asRanges(transcript, "exon")
try:
sequence = "".join(fasta.getSequence(contig, "+", exon[0], exon[1])
for exon in exons)
except KeyError:
continue
if strand == "-":
sequence = revcomp(sequence)
exon_counts = count_transcript(transcript, bam)
if exon_counts.sum() > 0:
yield (exon_counts, sequence)
# introns
intron_intervals = GTF.toIntronIntervals(transcript)
intron_counts = count_intervals(bam, intron_intervals, contig, strand)
if intron_counts.sum() == 0:
continue
for intron in intron_intervals:
seq = fasta.getSequence(contig, "+", intron[0], intron[1])
if strand == "-":
seq = revcomp(seq)
profile = intron_counts.loc[float(intron[0]):float(intron[1])]
profile.index = profile.index - intron[0]
if profile.sum() > 0:
yield (profile, seq)
def _get_counts_and_sequence(gtf_iterator, bam, fasta, seperate_UTRs=False):
'''Called by pentamer_enrichment. This function will return an iterator
that yeilds tuples of profiles accross transcripts or introns and the
sequence for which the profile is determined'''
for transcript in gtf_iterator:
transcript = [e for e in transcript if hasattr(e, "transcript_id")]
if len(transcript) == 0:
continue
contig, strand = transcript[0].contig, transcript[0].strand
# exons
exons = GTF.asRanges(transcript, "exon")
try:
sequence = "".join(
fasta.getSequence(contig, "+", exon[0], exon[1])
for exon in exons)
except KeyError:
continue
if strand == "-":
sequence = revcomp(sequence)
exon_counts = count_transcript(transcript, bam)
if exon_counts.sum() > 0:
yield (exon_counts, sequence)
# introns
intron_intervals = GTF.toIntronIntervals(transcript)
intron_counts = count_intervals(bam, intron_intervals, contig, strand)
if intron_counts.sum() == 0:
continue
for intron in intron_intervals:
seq = fasta.getSequence(contig, "+", intron[0], intron[1])
if strand == "-":
seq = revcomp(seq)
profile = intron_counts.loc[float(intron[0]):float(intron[1])]
profile.index = profile.index - intron[0]
if profile.sum() > 0:
yield (profile, seq)
def _get_profiles_and_conveter(gtf_iterator, bam):
for transcript in gtf_iterator:
transcript = [e for e in transcript if hasattr(e, "transcript_id")]
if len(transcript) == 0:
continue
gene_id = transcript[0].gene_id
#transcript_id = transcript[0].transcript_id
contig = transcript[0].contig
strand = transcript[0].strand
E.debug("Crunching gene: %s:"
% gene_id)
# exons
profile = count_transcript(transcript, bam)
if profile.sum() > 0:
converter = TranscriptCoordInterconverter(transcript)
yield (profile, converter, LiteExon(0, converter.length),
contig, strand)
# introns
intron_intervals = GTF.toIntronIntervals(transcript)
intron_counts = count_intervals(bam, intron_intervals,
contig, strand)
if intron_counts.sum() == 0:
continue
converter = TranscriptCoordInterconverter(transcript,
introns=True)
intron_counts.index = converter.genome2transcript(
intron_counts.index.values)
for intron in intron_intervals:
intron = (intron[0], intron[1] - 1)
intron = converter.genome2transcript(intron)
intron = sorted(intron)
intron = (intron[0], intron[1] + 1)
profile = intron_counts.loc[float(intron[0]):float(intron[1])]
if profile.sum() > 0:
yield (profile, converter, LiteExon(*intron), contig, strand)
def _get_profiles_and_conveter(gtf_iterator, bam):
for transcript in gtf_iterator:
transcript = [e for e in transcript if hasattr(e, "transcript_id")]
if len(transcript) == 0:
continue
gene_id = transcript[0].gene_id
#transcript_id = transcript[0].transcript_id
contig = transcript[0].contig
strand = transcript[0].strand
E.debug("Crunching gene: %s:" % gene_id)
# exons
profile = count_transcript(transcript, bam)
if profile.sum() > 0:
converter = TranscriptCoordInterconverter(transcript)
yield (profile, converter, LiteExon(0, converter.length), contig,
strand)
# introns
intron_intervals = GTF.toIntronIntervals(transcript)
intron_counts = count_intervals(bam, intron_intervals, contig, strand)
if intron_counts.sum() == 0:
continue
converter = TranscriptCoordInterconverter(transcript, introns=True)
intron_counts.index = converter.genome2transcript(
intron_counts.index.values)
for intron in intron_intervals:
intron = (intron[0], intron[1] - 1)
intron = converter.genome2transcript(intron)
intron = sorted(intron)
intron = (intron[0], intron[1] + 1)
profile = intron_counts.loc[float(intron[0]):float(intron[1])]
if profile.sum() > 0:
yield (profile, converter, LiteExon(*intron), contig, strand)
def processing_index(interval_iterator, bam, window_size=50):
'''Calculate the ratio of processed transcripts to non-processed
Parameters
----------
interval_iterator : CGAT.Bed or CGAT.GTF-like iterator
The iterator must yeild objects that have a start, end and strand
attribute. Processing index will be calculated around these.
bam : *_getter-like function
A getter function returned by the `make_getter` function, this will
be used to retrieve cross-link counts.
window_size : int, optional
How far up and downstream of the the processing site to consider.
Returns
-------
int
processing index averaged over all processing sites given.
Notes
-----
The proccessing index for G genes is defined as:
.. math::
pi = log_2( \frac{\sum_{i=1}^{G} N_i^{PM}}{\sum_{i=1}^{G} N_i^M})
after Baejen et al Mol Cell 5(55):745-757. However, Beaejen et al
normalise this number to the total number of genes, which seems
wrong to me. '''
n_pm = 0
n_m = 0
for site in interval_iterator:
if site.strand == "+":
pos = site.end
elif site.strand == "-":
pos = site.start
else:
raise ValueError(
"processing index not valid for unstranded cleavage points in "
"entry\n" + str(site) + "\n")
upstream_interval = (pos - window_size, pos)
downstream_interval = (pos, pos + window_size)
counts = count_intervals(bam, [upstream_interval, downstream_interval],
site.contig, site.strand)
# We are currently in genome cooridinates, not transcript
if site.strand == "+":
# pandas indexing is inclusive
n_up = counts.iloc[:pos - 1].sum()
n_down = counts.iloc[pos:].sum()
elif site.strand == "-":
n_up = counts.iloc[pos:].sum()
n_down = counts.iloc[:pos - 1].sum()
n_pm += n_down
n_m += n_up - n_down
pi = np.log2(float(n_pm) / float(max(1, n_m)))
return pi
def processing_index(interval_iterator, bam, window_size=50):
'''Calculate the ratio of processed transcripts to non-processed
Parameters
----------
interval_iterator : CGAT.Bed or CGAT.GTF-like iterator
The iterator must yeild objects that have a start, end and strand
attribute. Processing index will be calculated around these.
bam : *_getter-like function
A getter function returned by the `make_getter` function, this will
be used to retrieve cross-link counts.
window_size : int, optional
How far up and downstream of the the processing site to consider.
Returns
-------
int
processing index averaged over all processing sites given.
Notes
-----
The proccessing index for G genes is defined as:
.. math::
pi = log_2( \frac{\sum_{i=1}^{G} N_i^{PM}}{\sum_{i=1}^{G} N_i^M})
after Baejen et al Mol Cell 5(55):745-757. However, Beaejen et al
normalise this number to the total number of genes, which seems
wrong to me. '''
n_pm = 0
n_m = 0
for site in interval_iterator:
if site.strand == "+":
pos = site.end
elif site.strand == "-":
pos = site.start
else:
raise ValueError(
"processing index not valid for unstranded cleavage points in "
"entry\n" + str(site)+"\n")
upstream_interval = (pos - window_size, pos)
downstream_interval = (pos, pos + window_size)
counts = count_intervals(bam, [upstream_interval, downstream_interval],
site.contig, site.strand)
# We are currently in genome cooridinates, not transcript
if site.strand == "+":
# pandas indexing is inclusive
n_up = counts.iloc[:pos-1].sum()
n_down = counts.iloc[pos:].sum()
elif site.strand == "-":
n_up = counts.iloc[pos:].sum()
n_down = counts.iloc[:pos-1].sum()
n_pm += n_down
n_m += n_up-n_down
pi = np.log2(float(n_pm)/float(max(1, n_m)))
return pi