def dosage_generator(args, variant_mapping=None, weights=None):
if args.liftover:
logging.info("Acquiring liftover conversion")
liftover_chain = pyliftover.LiftOver(args.liftover)
liftover_conversion = lambda chr, pos: Genomics.lift(
liftover_chain, chr, pos, args.zero_based_positions)
else:
liftover_chain = None
liftover_conversion = None
whitelist = None
if variant_mapping and type(variant_mapping) == dict:
logging.info("Setting whitelist from mapping keys")
whitelist = set(variant_mapping.keys())
else:
logging.info("Setting whitelist from available models")
whitelist = set(weights.rsid)
d = None
if args.text_genotypes:
from metax.genotype import DosageGenotype
d = DosageGenotype.dosage_files_geno_lines(
args.text_genotypes,
variant_mapping=variant_mapping,
whitelist=whitelist,
skip_palindromic=args.skip_palindromic,
liftover_conversion=liftover_conversion)
elif args.bgen_genotypes:
from metax.genotype import BGENGenotype
d = BGENGenotype.bgen_files_geno_lines(
args.bgen_genotypes,
variant_mapping=variant_mapping,
force_colon=args.force_colon,
use_rsid=args.bgen_use_rsid,
whitelist=whitelist,
skip_palindromic=args.skip_palindromic)
elif args.vcf_genotypes:
from metax.genotype import CYVCF2Genotype
d = CYVCF2Genotype.vcf_files_geno_lines(
args.vcf_genotypes,
mode=args.vcf_mode,
variant_mapping=variant_mapping,
whitelist=whitelist,
skip_palindromic=args.skip_palindromic,
liftover_conversion=liftover_conversion)
if d is None:
raise Exceptions.InvalidArguments("unsupported genotype input")
if args.force_mapped_metadata:
d = Genotype.force_mapped_metadata(d, args.force_mapped_metadata)
return d
def get_variant_mapping(args, weights):
mapping = None
if len(args.variant_mapping):
if len(args.variant_mapping) == 3:
logging.info("Acquiring variant mapping")
mapping = KeyedDataSource.load_data(args.variant_mapping[0],
args.variant_mapping[1],
args.variant_mapping[2],
value_white_list=set(
weights.rsid))
# if args.variant_mapping[1] == "UKB":
# mapping = KeyedDataSource.load_data(args.variant_mapping[0], "variant", "panel_variant_id", value_white_list=set(weights.rsid))
# elif args.variant_mapping[1] == "RSID":
# mapping = KeyedDataSource.load_data(args.variant_mapping[0], "variant", "rsid", value_white_list=set(weights.rsid))
# elif args.variant_mapping[1] == "ID_TO_RSID":
# mapping = KeyedDataSource.load_data(args.variant_mapping[0], "id", "rsid", value_white_list=set(weights.rsid))
else:
raise Exceptions.InvalidArguments(
"Unsupported variant mapping argument")
elif len(args.on_the_fly_mapping):
checklist = set(weights.rsid)
if len(args.on_the_fly_mapping) > 0:
logging.info("Acquiring on-the-fly mapping")
if args.on_the_fly_mapping[0] == "METADATA":
if mapping:
_mapping = mapping # Python scope subtlety, they are not blocks like swift
mapping = lambda chromosome, position, ref_allele, alt_allele: Genomics.map_on_the_fly(
_mapping, args.on_the_fly_mapping[1], chromosome, position,
ref_allele, alt_allele)
else:
mapping = lambda chromosome, position, ref_allele, alt_allele: Genomics.coordinate_format(
checklist, args.on_the_fly_mapping[1], chromosome,
position, ref_allele, alt_allele)
else:
raise RuntimeError("Unsupported on_the_fly argument")
return mapping
def vcf_file_geno_lines(path,
mode="genotyped",
variant_mapping=None,
whitelist=None,
skip_palindromic=False,
liftover_conversion=None):
logging.log(9, "Processing vcf %s", path)
vcf_reader = VCF(path)
is_dict_mapping = variant_mapping is not None and type(
variant_mapping) == dict
for variant in vcf_reader:
chr = variant.CHROM
pos = variant.POS
variant_id = variant.ID
ref = variant.REF
alts = variant.ALT
if liftover_conversion:
chr_, pos_ = chr, pos
chr, pos = liftover_conversion(chr, pos)
if chr == "NA" or pos == "NA":
continue
if mode == "genotyped":
for a, alt in enumerate(alts):
if skip_palindromic and Genomics.is_palindromic(ref, alt):
continue
_varid, variant_id = Genomics.maybe_map_variant(
variant_id, chr, pos, ref, alt, variant_mapping,
is_dict_mapping)
if variant_id is None: continue
if whitelist and variant_id not in whitelist:
continue
d = []
for sample in variant.genotypes:
d_ = (sample[0] == a + 1) + (sample[1] == a + 1)
d.append(d_)
f = numpy.mean(numpy.array(d, dtype=numpy.int32)) / 2
yield (variant_id, chr, pos, ref, alt, f) + tuple(d)
elif mode == "imputed":
if len(alts) > 1:
logging.log(
"VCF imputed mode doesn't support multiple ALTs, skipping %s",
variant_id)
continue
alt = alts[0]
if skip_palindromic and Genomics.is_palindromic(ref, alt):
continue
_varid, variant_id = Genomics.maybe_map_variant(
variant_id, chr, pos, ref, alt, variant_mapping,
is_dict_mapping)
if variant_id is None: continue
if whitelist and variant_id not in whitelist:
continue
d = numpy.apply_along_axis(lambda x: x[0], 1, variant.format("DS"))
f = numpy.mean(numpy.array(d)) / 2
yield (variant_id, chr, pos, ref, alt, f) + tuple(d)
else:
raise RuntimeError("Unsupported vcf mode")