def get_rates_by_constraint(constraint, cache_dir, threshold=1e-4, ratio=1.0):
''' get mutation rates in and out of constrained regions
'''
rates = {'constrained': [], 'unconstrained': []}
mut_dict = load_mutation_rates()
ensembl = EnsemblRequest(cache_dir, 'grch37')
for tx_id, group in constraint.groupby('transcript'):
tx = construct_gene_object(ensembl, tx_id.split('.')[0])
sites = SiteRates(tx, mut_dict)
constrained_sites = get_constrained_positions(tx, group, threshold,
ratio)
cqs = [
'nonsense', 'missense', 'synonymous', 'splice_lof', 'splice_region'
]
gene_rates = get_gene_rates(tx, sites, cqs, constrained_sites)
# now add the gene rates to the larger list of all genes
for category in ['constrained', 'unconstrained']:
gene_rates[category]['symbol'] = list(group['gene'])[0]
gene_rates[category]['chrom'] = list(group['chr'])[0]
gene_rates[category]['length'] = tx.chrom_pos_to_cds(
tx.get_cds_end())['pos']
rates[category].append(gene_rates[category])
return rates
def annotate_constraint(data, constraint_path, threshold=1e-3, ratio=0.4):
''' annotate per-site rates by whether the site is under regional constraint
'''
# default to unconstrained
data['constrained'] = False
constraint = load_regional_constraint(constraint_path)
mut_dict = load_mutation_rates()
ensembl = EnsemblRequest(cache_dir, 'grch37')
modified = []
for symbol, group in data.groupby('symbol'):
if symbol not in set(constraint['gene']):
sites = set([])
else:
regional = constraint[constraint['gene'] == symbol]
tx_id = list(regional['transcript'])[0]
tx = construct_gene_object(ensembl, tx_id.split('.')[0])
sites = get_constrained_positions(tx, regional, threshold, ratio)
gene_constraint = group['constrained'].copy()
gene_constraint.loc[group['pos'].isin(sites)] = True
group['constrained'] = gene_constraint
modified.append(group)
return pandas.concat(modified)
def get_constrained_positions(ensembl, constraint, symbol, threshold=1e-4, ratio_threshold=1.0):
''' get the positions in the constrained regions
'''
regions = IntervalTree()
if symbol not in constraint:
return regions
data = constraint[symbol]
tx = construct_gene_object(ensembl, data['tx'])
for region in data['regions']:
p_value = chi2.sf(region['chisq'], df=1)
if p_value > threshold:
continue
if region['ratio'] > ratio_threshold:
continue
start, end = aa_to_chrom(tx, region['pos'])
if tx.get_strand() == '-':
start, end = end, start
regions[start:end + 1] = True
return regions
def test_construct_gene_object(self):
"""
"""
transcript_id = "ENST00000242577"
transcript = construct_gene_object(self.ensembl, transcript_id)
expected = self.set_transcript()
self.assertEqual(transcript, expected)
self.assertEqual(transcript.get_genomic_sequence(), expected.get_genomic_sequence())
self.assertEqual(transcript.get_cds_sequence(), expected.get_cds_sequence())
def test_construct_gene_object(self):
"""
"""
transcript_id = "ENST00000242577"
transcript = construct_gene_object(self.ensembl, transcript_id)
expected = self.set_transcript()
self.assertEqual(transcript, expected)
self.assertEqual(transcript.get_genomic_sequence(), expected.get_genomic_sequence())
self.assertEqual(transcript.get_cds_sequence(), expected.get_cds_sequence())
def get_mutation_rates(transcripts, mut_dict, ensembl):
""" determines mutation rates per functional category for transcripts
Args:
transcripts: list of transcript IDs for a gene
mut_dict: dictionary of local sequence context mutation rates
ensembl: EnsemblRequest object, to retrieve information from Ensembl.
Returns:
tuple of (rates, merged transcript, and transcript CDS length)
"""
rates = {
'missense': 0,
'nonsense': 0,
'splice_lof': 0,
'splice_region': 0,
'synonymous': 0
}
combined = None
for tx_id in transcripts:
try:
tx = construct_gene_object(ensembl, tx_id)
except ValueError:
continue
if len(tx.get_cds_sequence()) % 3 != 0:
raise ValueError("anomalous_coding_sequence")
# ignore mitochondrial genes
if tx.get_chrom() == "MT":
continue
sites = SiteRates(tx, mut_dict, masked_sites=combined)
combined = tx + combined
for cq in [
'missense', 'nonsense', 'splice_lof', 'splice_region',
'synonymous'
]:
rates[cq] += sites[cq].get_summed_rate()
if combined is None:
raise ValueError('no tx found')
length = combined.get_coding_distance(combined.get_cds_start(),
combined.get_cds_end())
return rates, combined, length
def get_mutation_rates(transcripts, mut_dict, ensembl):
""" determines mutation rates per functional category for transcripts
Args:
transcripts: list of transcript IDs for a gene
mut_dict: dictionary of local sequence context mutation rates
ensembl: EnsemblRequest object, to retrieve information from Ensembl.
Returns:
tuple of (rates, merged transcript, and transcript CDS length)
"""
rates = {'missense': 0, 'nonsense': 0, 'splice_lof': 0,
'splice_region': 0, 'synonymous': 0}
combined = None
for tx_id in transcripts:
try:
tx = construct_gene_object(ensembl, tx_id)
except ValueError:
continue
if len(tx.get_cds_sequence()) % 3 != 0:
raise ValueError("anomalous_coding_sequence")
# ignore mitochondrial genes
if tx.get_chrom() == "MT":
continue
sites = SiteRates(tx, mut_dict, masked_sites=combined)
combined = tx + combined
for cq in ['missense', 'nonsense', 'splice_lof', 'splice_region', 'synonymous']:
rates[cq] += sites[cq].get_summed_rate()
if combined is None:
raise ValueError('no tx found')
length = combined.get_coding_distance(combined.get_cds_end())['pos']
return rates, combined, length
def get_transcripts(symbol, ensembl):
''' get a list of Transcript objects for a gene
Args:
symbol: HGNC symbol for a gene
ensembl: EnsemblRequest object, to retrieve gene data with
Returns:
list of Transcript objects (see denovonear), sorted by size (longest
transcripts first)
'''
transcript_ids = get_transcript_ids(ensembl, symbol)
transcripts = []
for x in sorted(transcript_ids, key=transcript_ids.get, reverse=True):
try:
tx = construct_gene_object(ensembl, x)
transcripts.append(tx)
except ValueError:
continue
return transcripts
def get_mutation_rates(gene_id, transcripts, mut_dict, ensembl):
""" determines missense, nonsense and synonymous mutation rates for a gene
This can estimate a mutation rate from the union of transcripts for a gene.
This is a biased estimate of the mutation rate, where the mutation rate
estimates is biased towards the rate from the first-ranked transcripts,
which I prioritise by how many de novos they contain, and how long the
coding sequence is.
This isn't a problem when different transcripts have the same coding
sequence within their shared regions, as the rates will come outthe same,
but may differ two transcript share an overlapping region, but not in the
same frame, so that the sites that are missense, and nonsense will differ
between transcripts, and thus would produce different estimates of the
mutation rate.
Args:
gene_id: ID for the current gene (can be a transcript ID, if we are
examining single transcripts only, or can be a HGNC ID, if we are
examining the union of mutation rates from multiple transcripts for
a single gene).
transcripts: dictionary of transcripts for a gene, indexed by gene_id
mut_dict: dictionary of local sequence context mutation rates
ensembl: EnsemblRequest object, to retrieve information from Ensembl.
Returns:
tuple of (missense, nonsense, synonymous) mutation rates
"""
missense = 0
nonsense = 0
splice_lof = 0
splice_region = 0
synonymous = 0
combined_transcript = None
for transcript_id in transcripts[gene_id]:
# get the gene coordinates, sequence etc, but if the transcript is
# unusable (hence raises an error), simply move to the next transcript
try:
transcript = construct_gene_object(ensembl, transcript_id)
except ValueError:
continue
if len(transcript.get_cds_sequence()) % 3 != 0:
raise ValueError("anomalous_coding_sequence")
# ignore mitochondrial genes, since mitochondiral mutation rates differ
# from autosomal and allosomal mutation rates
if transcript.get_chrom() == "MT":
continue
if combined_transcript is None:
sites = SiteRates(transcript, mut_dict)
combined_transcript = transcript
else:
sites = SiteRates(transcript, mut_dict, masked_sites=combined_transcript)
combined_transcript += transcript
missense_rates = sites["missense"]
nonsense_rates = sites["nonsense"]
splice_lof_rates = sites["splice_lof"]
splice_region_rates = sites["splice_region"]
synonymous_rates = sites["synonymous"]
# if any sites have been sampled in the transcript, then add the
# cumulative probability from those sites to the approporiate
# mutation rate. Sometimes we won't have any sites for a transcript, as
# all the sites will have been captured in previous transcripts.
missense += missense_rates.get_summed_rate()
nonsense += nonsense_rates.get_summed_rate()
splice_lof += splice_lof_rates.get_summed_rate()
splice_region += splice_region_rates.get_summed_rate()
synonymous += synonymous_rates.get_summed_rate()
chrom = combined_transcript.get_chrom()
length = "NA"
if combined_transcript is not None:
length = combined_transcript.get_coding_distance(\
combined_transcript.get_cds_start(), combined_transcript.get_cds_end())
return (chrom, length, missense, nonsense, splice_lof, splice_region, synonymous)