def load_samples(args):
s = None
if args.text_sample_ids:
if len(args.text_sample_ids) == 1:
s = pandas.read_table(args.text_sample_ids[0],
header=None,
names=["FID", "IID"])
elif args.text_sample_ids[1] == "UKB":
k = pandas.read_table(args.text_sample_ids[0], sep=" ")
k = k[k.sex != "D"].reset_index(drop=True)
s = k[["ID_1", "ID_2"]].rename(columns={
"ID_1": "FID",
"ID_2": "IID"
})
elif args.vcf_genotypes:
from metax.genotype import CYVCF2Genotype
s = CYVCF2Genotype.get_samples(args.vcf_genotypes[0])
elif args.bgen_genotypes:
from metax.genotype import BGENGenotype
s = BGENGenotype.get_samples(args.bgen_genotypes[0])
elif args.generate_sample_ids:
s = ["ID_{}".format(x) for x in range(0, args.generate_sample_ids)]
s = [(x, x) for x in s]
s = pandas.DataFrame(data=s, columns=["FID", "IID"])
if s is None:
raise Exceptions.InvalidArguments("Unsupported samples argument")
return s
def _prepare_phenotype(context):
logging.info("Accquiring phenotype")
context.pheno = _pheno_from_file_and_column(context.args.input_phenos_file, context.args.input_phenos_column)
if context.args.mode == MTPMode.K_LOGISTIC:
try:
v = set([str(float(x)) for x in context.pheno])
if not v.issubset({'0.0', '1.0', 'nan'}):
raise Exceptions.InvalidArguments("Logistic regression was asked but phenotype is not binomial")
except:
raise Exceptions.InvalidArguments("Logistic regression: could not parse phenotype")
context.mode = context.args.mode
if context.args.covariates_file and context.args.covariates:
context.mode = MTPMode.K_LINEAR
logging.info("Acquiring covariates")
context.covariates = _get_covariates(context.args)
logging.info("Replacing phenotype with residuals")
context.pheno = _get_residual(context.pheno, context.covariates)
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 prepare_prediction(args, extra, samples):
logging.info("Preparing prediction")
results = None
if len(args.prediction_output) < 2:
from metax.predixcan.Utilities import BasicPredictionRepository
results = BasicPredictionRepository(samples, extra,
args.prediction_output[0])
else:
if args.prediction_output[1] == "HDF5":
from metax.predixcan.Utilities import HDF5PredictionRepository
results = HDF5PredictionRepository(samples, extra,
args.prediction_output[0])
else:
raise Exceptions.InvalidArguments(
"Unsupported output specification")
return results
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 validate(args):
if (args.gwas_file and args.gwas_folder) or (not args.gwas_file
and not args.gwas_folder):
raise Exceptions.InvalidArguments(
"Provide either (--gwas_file) or (--gwas_folder [--gwas_file_pattern])"
)
def validate(args):
if not args.gwas_folder:
raise Exceptions.InvalidArguments(
"You need to provide an input folder containing GWAS files")
def _check_args(args):
if not args.mode in MTPMode.K_MODES:
raise Exceptions.InvalidArguments("Invalid mode")
