validate AD length (#1058)

* validate AD length

* change error message language

* handle empties

* format

* ruff

Author: bpblanken

v03_pipeline/lib/misc/validation.py

@@ -48,6 +48,26 @@ def validate_allele_type(
         raise SeqrValidationError(msg)
 
 
+def validate_allele_depth_length(
+    mt: hl.MatrixTable,
+    reference_genome: ReferenceGenome,
+    dataset_type: DatasetType,
+    **_: Any,
+) -> None:
+    ht = mt.select_rows(
+        found_ad_lengths=hl.agg.collect_as_set(hl.len(mt.AD)).remove(
+            hl.missing(hl.tint32),
+        ),
+    ).rows()
+    ht = ht.filter(
+        hl.len(ht.found_ad_lengths) > 1,
+    )
+    if ht.count() > 0:
+        variant_format = dataset_type.table_key_format_fn(reference_genome)
+        msg = f'Found variants with unequal Allele Depth array lengths over samples (first 10, if applicable): {({variant_format(v): v.found_ad_lengths for v in ht.take(10)})}'
+        raise SeqrValidationError(msg)
+
+
 def validate_no_duplicate_variants(
     t: hl.Table | hl.MatrixTable,
     reference_genome: ReferenceGenome,

v03_pipeline/lib/misc/validation_test.py

@@ -4,6 +4,7 @@
 
 from v03_pipeline.lib.misc.validation import (
     SeqrValidationError,
+    validate_allele_depth_length,
     validate_allele_type,
     validate_expected_contig_frequency,
     validate_imported_field_types,
@@ -122,6 +123,63 @@ def test_validate_allele_type(self) -> None:
             DatasetType.SNV_INDEL,
         )
 
+    def test_validate_allele_depth_length(self) -> None:
+        mt = (
+            hl.MatrixTable.from_parts(
+                rows={
+                    'locus': [
+                        hl.Locus(
+                            contig='chr1',
+                            position=1,
+                            reference_genome='GRCh38',
+                        ),
+                        hl.Locus(
+                            contig='chr1',
+                            position=2,
+                            reference_genome='GRCh38',
+                        ),
+                        hl.Locus(
+                            contig='chr1',
+                            position=3,
+                            reference_genome='GRCh38',
+                        ),
+                        hl.Locus(
+                            contig='chr1',
+                            position=4,
+                            reference_genome='GRCh38',
+                        ),
+                    ],
+                    'alleles': [
+                        ['A', 'T'],
+                        # NB: star alleles should pass through this validation just fine,
+                        # but are eventually filtered out upstream.
+                        ['A', 'TC', 'TG'],
+                        ['A', 'TTT'],
+                        ['A', 'CCC'],
+                    ],
+                },
+                cols={'s': ['sample_1', 'sample_2']},
+                entries={
+                    'AD': [
+                        [[1, 0], [1, 0]],
+                        [[1], [1, 0, 1]],
+                        [[1, 0], [1]],
+                        [[1, 0], [1, 0]],
+                    ],
+                },
+            )
+            .key_rows_by('locus', 'alleles')
+            .key_cols_by('s')
+        )
+        self.assertRaisesRegex(
+            SeqrValidationError,
+            "Found variants with unequal Allele Depth array lengths over samples \\(first 10, if applicable\\): \\{'1-2-A-TC-TG': \\{1, 3\\}, '1-3-A-TTT': \\{1, 2\\}\\}",
+            validate_allele_depth_length,
+            mt,
+            ReferenceGenome.GRCh38,
+            DatasetType.SNV_INDEL,
+        )
+
     def test_validate_imputed_sex_ploidy(self) -> None:
         female_sample = 'HG00731_1'
         male_sample_1 = 'HG00732_1'

v03_pipeline/lib/model/dataset_type.py

@@ -36,7 +36,7 @@ def table_key_format_fn(
         if self in {DatasetType.GCNV, DatasetType.SV}:
             return lambda s: s.variant_id
         return (
-            lambda s: f'{s.locus.contig if reference_genome == ReferenceGenome.GRCh37 else s.locus.contig.replace("chr", "")}-{s.locus.position}-{s.alleles[0]}-{s.alleles[1]}'
+            lambda s: f'{s.locus.contig if reference_genome == ReferenceGenome.GRCh37 else s.locus.contig.replace("chr", "")}-{s.locus.position}-{"-".join(s.alleles)}'
         )
 
     @property

v03_pipeline/lib/tasks/validate_callset.py

@@ -4,6 +4,7 @@
 
 from v03_pipeline.lib.misc.validation import (
     SeqrValidationError,
+    validate_allele_depth_length,
     validate_allele_type,
     validate_expected_contig_frequency,
     validate_imputed_sex_ploidy,
@@ -123,6 +124,7 @@ def update_table(self, mt: hl.MatrixTable) -> hl.MatrixTable:
             )
         validation_dependencies = self.get_validation_dependencies()
         for validation_f in [
+            validate_allele_depth_length,
             validate_allele_type,
             validate_imputed_sex_ploidy,
             validate_no_duplicate_variants,