def label_variants(variants, truth_variants, ref):
"""Assigns genotypes to each variant to best match truth_variants.
See the module-level documentation for general information on how this
algorithm works.
Args:
variants: list[nucleus.protos.Variant]. A list of candidate variants, in
coordinate-sorted order, all on the same chromosome.
truth_variants: list[nucleus.protos.Variant]. A list of truth variants, in
coordinate-sorted order, for the same interval on the genome as variants.
ref: ReferenceRegion. Used to get reference bases for variants. Must cover
at least the span of the variants.
Returns:
A list of new variants, copied from variants, but with their
call[0].genotype field assigned values for the optimal labeling of variants.
Raises:
ValueError: If any inputs are malformed.
"""
variants = list(variants)
truth_variants = list(truth_variants)
if _DEBUG_PRINTING_IS_ENABLED:
print_variants('candidates', variants)
print_variants('truth', truth_variants)
if not variant_utils.variants_are_sorted(variants):
raise ValueError('Variants are not sorted', variants)
if not variant_utils.variants_are_sorted(truth_variants):
raise ValueError('truth_variants are not sorted', truth_variants)
truth_haplotypes = deduplicate_haplotypes(
enumerate_all_possible_haplotypes(truth_variants, ref,
EnumerationType.TRUTH))
# redacted
variant_haplotypes = enumerate_all_possible_haplotypes(
variants, ref, EnumerationType.CANDIDATES)
found = []
for vh, vgt in variant_haplotypes:
for th, tgt in truth_haplotypes:
if th == vh:
found.append(
LabelerMatch(
haplotypes=th,
variants=variants,
matched_variant_genotypes=vgt,
truth_variants=truth_variants,
matched_truth_genotypes=tgt))
if not found:
return None
else:
best = select_best_match(found)
return best.variants_with_assigned_genotypes()
def label_variants(variants, truth_variants, ref):
"""Assigns genotypes to each variant to best match truth_variants.
See the module-level documentation for general information on how this
algorithm works.
Args:
variants: list[nucleus.protos.Variant]. A list of candidate variants, in
coordinate-sorted order, all on the same chromosome.
truth_variants: list[nucleus.protos.Variant]. A list of truth variants, in
coordinate-sorted order, for the same interval on the genome as variants.
ref: ReferenceRegion. Used to get reference bases for variants. Must cover
at least the span of the variants.
Returns:
A list of new variants, copied from variants, but with their
call[0].genotype field assigned values for the optimal labeling of variants.
Raises:
ValueError: If any inputs are malformed.
"""
variants = list(variants)
truth_variants = list(truth_variants)
if _DEBUG_PRINTING_IS_ENABLED:
print_variants('candidates', variants)
print_variants('truth', truth_variants)
if not variant_utils.variants_are_sorted(variants):
raise ValueError('Variants are not sorted', variants)
if not variant_utils.variants_are_sorted(truth_variants):
raise ValueError('truth_variants are not sorted', truth_variants)
truth_haplotypes = deduplicate_haplotypes(
enumerate_all_possible_haplotypes(truth_variants, ref,
EnumerationType.TRUTH))
# redacted
variant_haplotypes = enumerate_all_possible_haplotypes(
variants, ref, EnumerationType.CANDIDATES)
found = []
for vh, vgt in variant_haplotypes:
for th, tgt in truth_haplotypes:
if th == vh:
found.append(
LabelerMatch(haplotypes=th,
variants=variants,
matched_variant_genotypes=vgt,
truth_variants=truth_variants,
matched_truth_genotypes=tgt))
if not found:
return None
else:
best = select_best_match(found)
return best.variants_with_assigned_genotypes()
def _raise_if_not_sorted_or_not_on_same_chromosome(variants):
"""Raises a ValueError if variants isn't sorted on the same chromosome."""
if not variant_utils.variants_are_sorted(variants):
raise ValueError('Variants must be sorted', variants)
for v in variants[1:]:
if variants[0].reference_name != v.reference_name:
raise ValueError(
'Variants (v1={}, v2={}) not on the same chromosome'.format(
v.reference_name, variants[0].reference_name))
def test_sorted_variants(self, sorted_variants):
for permutation in itertools.permutations(
sorted_variants, r=len(sorted_variants)):
# Check that sorting the permutations produced sorted.
self.assertEqual(
variant_utils.sorted_variants(permutation), sorted_variants)
# Check that variants_are_sorted() is correct, which we detect if
# the range_tuples of permutation == the range_tuples of sorted_variants.
def _range_tuples(variants):
return [variant_utils.variant_range_tuple(v) for v in variants]
self.assertEqual(
variant_utils.variants_are_sorted(permutation),
_range_tuples(permutation) == _range_tuples(sorted_variants))
def test_sorted_variants(self, sorted_variants):
for permutation in itertools.permutations(
sorted_variants, r=len(sorted_variants)):
# Check that sorting the permutations produced sorted.
self.assertEqual(
variant_utils.sorted_variants(permutation), sorted_variants)
# Check that variants_are_sorted() is correct, which we detect if
# the range_tuples of permutation == the range_tuples of sorted_variants.
def _range_tuples(variants):
return [variant_utils.variant_range_tuple(v) for v in variants]
self.assertEqual(
variant_utils.variants_are_sorted(permutation),
_range_tuples(permutation) == _range_tuples(sorted_variants))
def find_best_matching_haplotypes(candidates, truths, ref):
"""Assigns genotypes to each variant to best match truths.
See the module-level documentation for general information on how this
algorithm works.
Args:
candidates: list[nucleus.protos.Variant]. A list of candidate variants, in
coordinate-sorted order, all on the same chromosome.
truths: list[nucleus.protos.Variant]. A list of truth variants, in
coordinate-sorted order, for the same interval on the genome as variants.
ref: ReferenceRegion. Used to get reference bases for variants. Must cover
at least the span of the variants.
Returns:
A HaplotypeMatch object describing the best assignment of genotypes between
the candidates and truth_variants, or None, if no consistent assignment can
be found.
Raises:
ValueError: If any inputs are malformed.
"""
candidates = list(candidates)
truths = list(truths)
if _DEBUG_PRINTING_IS_ENABLED:
_log_variants('candidates', candidates)
_log_variants('truth', truths)
if not variant_utils.variants_are_sorted(candidates):
raise ValueError('candidates are not sorted', candidates)
if not variant_utils.variants_are_sorted(truths):
raise ValueError('truths are not sorted', truths)
def _hom_ref_enum_if_empty(list_of_variants, non_empty_enum):
"""If list_of_variants is empty, use a ONLY_HOM_REF enum for speed."""
return non_empty_enum if list_of_variants else EnumerationType.ONLY_HOM_REF
truth_haplotypes = deduplicate_haplotypes(
enumerate_all_possible_haplotypes(
truths, ref,
_hom_ref_enum_if_empty(candidates, EnumerationType.TRUTH)))
# Note, it may be worth deduplicating these haplotypes as well.
variant_haplotypes = enumerate_all_possible_haplotypes(
candidates, ref,
_hom_ref_enum_if_empty(truths, EnumerationType.CANDIDATES))
found = []
for vh, vgt in variant_haplotypes:
for th, tgt in truth_haplotypes:
if th == vh:
found.append(
HaplotypeMatch(haplotypes=th,
candidates=candidates,
candidate_genotypes=vgt,
truths=truths,
truth_genotypes=tgt))
if not found:
return None
else:
return select_best_haplotype_match(found)