run validation on sv get_ht

Author: jklugherz

v03_pipeline/lib/annotations/sv.py

@@ -180,6 +180,7 @@ def gnomad_svs(
     gnomad_svs_ht: hl.Table,
     **_: Any,
 ) -> hl.Expression:
+    gnomad_svs_ht = gnomad_svs_ht.drop('locus', 'alleles')
     return gnomad_svs_ht.annotate(
         ID=gnomad_svs_ht.KEY,
     )[ht['info.GNOMAD_V4.1_TRUTH_VID']]

v03_pipeline/lib/misc/validation.py

@@ -31,9 +31,6 @@ def validate_allele_type(
     **_: Any,
 ) -> None:
     ht = t.rows() if isinstance(t, hl.MatrixTable) else t
-    if not hasattr(ht, 'alleles'):
-        return
-
     ht = ht.filter(
         dataset_type.invalid_allele_types.contains(
             hl.numeric_allele_type(ht.alleles[0], ht.alleles[1]),

v03_pipeline/lib/reference_datasets/gnomad_svs.py

@@ -14,5 +14,4 @@ def get_ht(path: str, reference_genome: ReferenceGenome) -> hl.Table:
         N_HET=ht.info.N_HET,
         N_HOMREF=ht.info.N_HOMREF,
     )
-    ht = ht.key_by('KEY')
-    return ht.drop('locus', 'alleles')
+    return ht.key_by('KEY')

v03_pipeline/lib/reference_datasets/gnomad_svs_test.py

@@ -21,6 +21,12 @@ def test_gnomad_svs(self, mock_vcf_to_ht):
             [
                 hl.Struct(
                     KEY='gnomAD-SV_v3_BND_chr1_1a45f73a',
+                    locus=hl.Locus(
+                        contig='chr1',
+                        position=10434,
+                        reference_genome=ReferenceGenome.GRCh38,
+                    ),
+                    alleles=['N', '<BND>'],
                     AF=0.11413399875164032,
                     AC=8474,
                     AN=74246,
@@ -29,6 +35,12 @@ def test_gnomad_svs(self, mock_vcf_to_ht):
                 ),
                 hl.Struct(
                     KEY='gnomAD-SV_v3_BND_chr1_3fa36917',
+                    locus=hl.Locus(
+                        contig='chr1',
+                        position=10440,
+                        reference_genome=ReferenceGenome.GRCh38,
+                    ),
+                    alleles=['N', '<BND>'],
                     AF=0.004201000090688467,
                     AC=466,
                     AN=110936,
@@ -37,6 +49,12 @@ def test_gnomad_svs(self, mock_vcf_to_ht):
                 ),
                 hl.Struct(
                     KEY='gnomAD-SV_v3_BND_chr1_7bbf34b5',
+                    locus=hl.Locus(
+                        contig='chr1',
+                        position=10464,
+                        reference_genome=ReferenceGenome.GRCh38,
+                    ),
+                    alleles=['N', '<BND>'],
                     AF=0.03698499873280525,
                     AC=3119,
                     AN=84332,
@@ -45,6 +63,12 @@ def test_gnomad_svs(self, mock_vcf_to_ht):
                 ),
                 hl.Struct(
                     KEY='gnomAD-SV_v3_BND_chr1_933a2971',
+                    locus=hl.Locus(
+                        contig='chr1',
+                        position=10450,
+                        reference_genome=ReferenceGenome.GRCh38,
+                    ),
+                    alleles=['N', '<BND>'],
                     AF=0.3238990008831024,
                     AC=21766,
                     AN=67200,
@@ -53,6 +77,12 @@ def test_gnomad_svs(self, mock_vcf_to_ht):
                 ),
                 hl.Struct(
                     KEY='gnomAD-SV_v3_DUP_chr1_01c2781c',
+                    locus=hl.Locus(
+                        contig='chr1',
+                        position=10000,
+                        reference_genome=ReferenceGenome.GRCh38,
+                    ),
+                    alleles=['N', '<DUP>'],
                     AF=0.0019970000721514225,
                     AC=139,
                     AN=69594,

v03_pipeline/lib/reference_datasets/reference_dataset.py

@@ -151,7 +151,7 @@ def get_ht(
         ht = filter_contigs(ht, reference_genome)
         for dataset_type in self.dataset_types(reference_genome):
             validate_allele_type(ht, dataset_type)
-        validate_no_duplicate_variants(ht, reference_genome, DatasetType.SNV_INDEL)
+            validate_no_duplicate_variants(ht, reference_genome, dataset_type)
         # NB: we do not filter with "filter" here
         # ReferenceDatasets are DatasetType agnostic and that
         # filter is only used at annotation time.