Merge pull request #702 from broadinstitute/benb/reference_genome_in_vep_config

Add reference genome check to VEP.

Author: bpblanken

v03_pipeline/lib/annotations/fields_test.py

@@ -31,13 +31,16 @@ def setUp(self) -> None:
             ),
         )
 
+    @patch('v03_pipeline.lib.vep.validate_vep_config_reference_genome')
     @patch('v03_pipeline.lib.vep.hl.vep')
-    def test_get_formatting_fields(self, mock_vep: Mock) -> None:
+    def test_get_formatting_fields(self, mock_vep: Mock, mock_validate: Mock) -> None:
         ht = hl.read_table(TEST_COMBINED_1)
         mock_vep.return_value = ht.annotate(vep=MOCK_VEP_DATA)
+        mock_validate.return_value = None
         ht = run_vep(
             ht,
             DatasetType.SNV_INDEL,
+            ReferenceGenome.GRCh38,
             None,
         )
         ht = ht.annotate(rsid='abcd')

v03_pipeline/lib/annotations/shared_test.py

@@ -10,8 +10,13 @@
 
 
 class SharedAnnotationsTest(unittest.TestCase):
+    @patch('v03_pipeline.lib.vep.validate_vep_config_reference_genome')
     @patch('v03_pipeline.lib.vep.hl.vep')
-    def test_sorted_transcript_consequences(self, mock_vep: Mock) -> None:
+    def test_sorted_transcript_consequences(
+        self,
+        mock_vep: Mock,
+        mock_validate: Mock,
+    ) -> None:
         ht = hl.Table.parallelize(
             [
                 {
@@ -30,9 +35,11 @@ def test_sorted_transcript_consequences(self, mock_vep: Mock) -> None:
             key=['locus', 'alleles'],
         )
         mock_vep.return_value = ht.annotate(vep=MOCK_VEP_DATA)
+        mock_validate.return_value = None
         ht = run_vep(
             ht,
             DatasetType.SNV_INDEL,
+            ReferenceGenome.GRCh38,
             None,
         )
         ht = ht.select(

v03_pipeline/lib/tasks/update_variant_annotations_table_with_new_samples.py

@@ -215,6 +215,7 @@ def update_table(self, ht: hl.Table) -> hl.Table:
         new_variants_ht = run_vep(
             new_variants_ht,
             self.dataset_type,
+            self.reference_genome,
             self.vep_config_json_path,
         )
 

v03_pipeline/lib/tasks/update_variant_annotations_table_with_new_samples_test.py

@@ -186,14 +186,17 @@ def test_missing_interval_reference(self, mock_update_rdc_task) -> None:
     )
     @patch.object(ReferenceGenome, 'standard_contigs', new_callable=PropertyMock)
     @patch('v03_pipeline.lib.vep.hl.vep')
+    @patch('v03_pipeline.lib.vep.validate_vep_config_reference_genome')
     def test_multiple_update_vat(
         self,
+        mock_vep_validate: Mock,
         mock_vep: Mock,
         mock_standard_contigs: Mock,
         mock_update_rdc_task: Mock,
     ) -> None:
         mock_update_rdc_task.return_value = MockCompleteTask()
         mock_vep.side_effect = lambda ht, **_: ht.annotate(vep=MOCK_VEP_DATA)
+        mock_vep_validate.return_value = None
         mock_standard_contigs.return_value = {'chr1'}
         # This creates a mock validation table with 1 coding and 1 non-coding variant
         # explicitly chosen from the VCF.
@@ -504,13 +507,16 @@ def test_multiple_update_vat(
         )
 
     @patch('v03_pipeline.lib.vep.hl.vep')
+    @patch('v03_pipeline.lib.vep.validate_vep_config_reference_genome')
     def test_update_vat_grch37(
         self,
+        mock_vep_validate: Mock,
         mock_vep: Mock,
         mock_update_rdc_task: Mock,
     ) -> None:
         mock_update_rdc_task.return_value = MockCompleteTask()
         mock_vep.side_effect = lambda ht, **_: ht.annotate(vep=MOCK_VEP_DATA)
+        mock_vep_validate.return_value = None
         worker = luigi.worker.Worker()
         uvatwns_task = UpdateVariantAnnotationsTableWithNewSamplesTask(
             reference_genome=ReferenceGenome.GRCh37,
@@ -551,8 +557,10 @@ def test_update_vat_grch37(
 
     @patch('v03_pipeline.lib.model.reference_dataset_collection.Env')
     @patch('v03_pipeline.lib.vep.hl.vep')
+    @patch('v03_pipeline.lib.vep.validate_vep_config_reference_genome')
     def test_update_vat_without_accessing_private_datasets(
         self,
+        mock_vep_validate: Mock,
         mock_vep: Mock,
         mock_rdc_env: Mock,
         mock_update_rdc_task: Mock,
@@ -567,6 +575,7 @@ def test_update_vat_without_accessing_private_datasets(
         )
         mock_rdc_env.ACCESS_PRIVATE_REFERENCE_DATASETS = False
         mock_vep.side_effect = lambda ht, **_: ht.annotate(vep=MOCK_VEP_DATA)
+        mock_vep_validate.return_value = None
         worker = luigi.worker.Worker()
         uvatwns_task = UpdateVariantAnnotationsTableWithNewSamplesTask(
             reference_genome=ReferenceGenome.GRCh38,

v03_pipeline/lib/vep.py

@@ -1,20 +1,29 @@
 import hail as hl
 
-from v03_pipeline.lib.model import DatasetType
+from v03_pipeline.lib.model import DatasetType, ReferenceGenome
+
+
+def validate_vep_config_reference_genome(reference_genome, config: str) -> None:
+    with open(config) as f:
+        if reference_genome.value not in f.read():
+            msg = f'Vep config does not match supplied reference genome {reference_genome.value}'
+            raise ValueError(msg)
 
 
 def run_vep(
     ht: hl.Table,
     dataset_type: DatasetType,
+    reference_genome: ReferenceGenome,
     vep_config_json_path: str | None,
 ) -> hl.Table:
     if not dataset_type.veppable:
         return ht
     config = (
         vep_config_json_path
         if vep_config_json_path is not None
-        else 'file:///vep_data/vep-gcloud.json'
+        else f'file:///vep_data/vep-{reference_genome.value}-gcloud.json'
     )
+    validate_vep_config_reference_genome(reference_genome, config)
     return hl.vep(
         ht,
         config=config,