def test_vcf_vds_combiner_equivalence():
import hail.experimental.vcf_combiner.vcf_combiner as vcf
import hail.vds.combiner as vds
_paths = ['gvcfs/HG00096.g.vcf.gz', 'gvcfs/HG00268.g.vcf.gz']
paths = [resource(p) for p in _paths]
parts = [
hl.Interval(start=hl.Struct(locus=hl.Locus('chr20', 17821257, reference_genome='GRCh38')),
end=hl.Struct(locus=hl.Locus('chr20', 18708366, reference_genome='GRCh38')),
includes_end=True),
hl.Interval(start=hl.Struct(locus=hl.Locus('chr20', 18708367, reference_genome='GRCh38')),
end=hl.Struct(locus=hl.Locus('chr20', 19776611, reference_genome='GRCh38')),
includes_end=True),
hl.Interval(start=hl.Struct(locus=hl.Locus('chr20', 19776612, reference_genome='GRCh38')),
end=hl.Struct(locus=hl.Locus('chr20', 21144633, reference_genome='GRCh38')),
includes_end=True)
]
vcfs = [mt.annotate_rows(info=mt.info.annotate(
MQ_DP=hl.missing(hl.tint32),
VarDP=hl.missing(hl.tint32),
QUALapprox=hl.missing(hl.tint32)))
for mt in hl.import_gvcfs(paths, parts, reference_genome='GRCh38',
array_elements_required=False)]
entry_to_keep = defined_entry_fields(vcfs[0].filter_rows(hl.is_defined(vcfs[0].info.END)), 100_000) - {'GT', 'PGT', 'PL'}
vds = vds.combine_variant_datasets([vds.transform_gvcf(mt, reference_entry_fields_to_keep=entry_to_keep) for mt in vcfs])
smt = vcf.combine_gvcfs([vcf.transform_gvcf(mt) for mt in vcfs])
smt_from_vds = hl.vds.to_merged_sparse_mt(vds).drop('RGQ')
smt = smt.select_entries(*smt_from_vds.entry) # harmonize fields and order
smt = smt.key_rows_by('locus', 'alleles')
assert smt._same(smt_from_vds)
def compile_2k_merge(path):
vcf = setup(path)
vcfs = [vc_all.transform_gvcf(vcf)] * COMBINE_GVCF_MAX
combined = [vc_all.combine_gvcfs(vcfs)] * 20
with TemporaryDirectory() as tmpdir:
hl.experimental.write_matrix_tables(combined,
os.path.join(
tmpdir,
'combiner-multi-write'),
overwrite=True)
def compile_2k_merge(path):
flagname = 'no_ir_logging'
prev_flag_value = hl._get_flags(flagname).get(flagname)
try:
hl._set_flags(**{flagname: '1'})
vcf = setup(path)
vcfs = [vc_all.transform_gvcf(vcf)] * COMBINE_GVCF_MAX
combined = [vc_all.combine_gvcfs(vcfs)] * 20
with TemporaryDirectory() as tmpdir:
hl.experimental.write_matrix_tables(combined,
os.path.join(
tmpdir,
'combiner-multi-write'),
overwrite=True)
finally:
hl._set_flags(**{flagname: prev_flag_value})
def main(args):
hl.init(default_reference='GRCh38')
hgdp_inputs = []
tgp_inputs = []
with hl.hadoop_open(
'gs://hgdp_tgp/misc/tgp_plus_hgdp_30x_reblocked_gvcfs.txt',
'r') as f:
for line in f:
line = line.strip()
hgdp_inputs.append(line) if 'HGDP' in line else tgp_inputs.append(
line)
temp_bucket = 'gs://gnomad-tmp/tgp_hgdp'
if args.get_sample_names:
get_sample_names_from_list_of_files(tgp_inputs,
get_samples_path('tgp'))
get_sample_names_from_list_of_files(hgdp_inputs,
get_samples_path('hgdp'))
if args.create_sparse_mt:
sample_names = get_sample_list_in_order(get_samples_path('tgp'),
tgp_inputs)
hl.experimental.run_combiner(tgp_inputs,
out_file=get_reference_mt_path(
'tgp', sparse=True),
tmp_path=temp_bucket,
overwrite=args.overwrite,
header=get_header_path('tgp'),
sample_names=sample_names,
use_genome_default_intervals=True)
sample_names = get_sample_list_in_order(get_samples_path('hgdp'),
hgdp_inputs)
hl.experimental.run_combiner(hgdp_inputs,
out_file=get_reference_mt_path(
'hgdp', sparse=True),
tmp_path=temp_bucket,
overwrite=args.overwrite,
header=get_header_path('hgdp'),
sample_names=sample_names,
use_genome_default_intervals=True)
tgp_mt = hl.read_matrix_table(get_reference_mt_path('tgp',
sparse=True))
tgp_mt = tgp_mt.annotate_entries(
gvcf_info=tgp_mt.gvcf_info.drop('MQ0', 'VariantType')).drop(
'AB', 'MQ0')
hgdp_mt = hl.read_matrix_table(
get_reference_mt_path('hgdp', sparse=True))
hgdp_mt = hgdp_mt.annotate_entries(gvcf_info=hgdp_mt.gvcf_info.select(
*tgp_mt.gvcf_info))
mt = combine_gvcfs([tgp_mt, hgdp_mt])
mt.write(get_reference_mt_path(sparse=True), overwrite=args.overwrite)
if args.densify_mt:
mt = hl.read_matrix_table(get_reference_mt_path(
sparse=True)).key_rows_by('locus', 'alleles')
mt = hl.experimental.densify(hl.experimental.sparse_split_multi(mt))
mt = mt.filter_rows(hl.len(mt.alleles) > 1)
mt.naive_coalesce(5000).write(get_reference_mt_path(), args.overwrite)
mt = hl.read_matrix_table(get_reference_mt_path()).drop('gvcf_info')
hl.export_vcf(mt,
get_reference_mt_path(extension='vcf.bgz'),
parallel='header_per_shard')
def python_only_10k_combine(path):
vcf = setup(path)
mt = vc_all.transform_gvcf(vcf)
mts = [mt] * 10_000
_ = [vc_all.combine_gvcfs(mts) for mts in chunks(mts, COMBINE_GVCF_MAX)]
def combine_variant_datasets(vdss: List[VariantDataset]) -> VariantDataset:
reference = combine_references([vds.reference_data for vds in vdss])
no_variant_key = [vds.variant_data.key_rows_by('locus') for vds in vdss]
variants = combine_gvcfs(no_variant_key)
return VariantDataset(reference, variants._key_rows_by_assert_sorted('locus', 'alleles'))
