feat(pipelines): update v4 genes to annotate low coverage transcripts

Author: rileyhgrant

data-pipeline/src/data_pipeline/data_types/gene.py

@@ -164,6 +164,24 @@ def collect_transcript_exons(transcript_exons):
     return exons
 
 
+def annotate_gene_models_with_low_coverage_flag(genes_path, low_coverage_tsv_path):
+    low_coverage_flag_name = "v4_low_coverage_trancript"
+
+    genes_ht = hl.read_table(genes_path)
+    tsv_ht = hl.import_table(low_coverage_tsv_path)
+    tsv_ht = tsv_ht.key_by("transcript_id")
+
+    genes_ht = genes_ht.annotate(
+        flags=hl.if_else(
+            hl.is_defined(tsv_ht[genes_ht.canonical_transcript_id]),
+            hl.or_else(genes_ht.flags, hl.empty_set(hl.tstr)).add(low_coverage_flag_name),
+            genes_ht.flags,
+        )
+    )
+
+    return genes_ht
+
+
 ###############################################
 # Main                                        #
 ###############################################

data-pipeline/src/data_pipeline/pipelines/genes.py

@@ -32,7 +32,7 @@
     prepare_heterozygous_variant_cooccurrence_counts,
     prepare_homozygous_variant_cooccurrence_counts,
 )
-from data_pipeline.data_types.gene import reject_par_y_genes
+from data_pipeline.data_types.gene import reject_par_y_genes, annotate_gene_models_with_low_coverage_flag
 
 from data_pipeline.datasets.gnomad_v4.gnomad_v4_constraint import (
     prepare_gnomad_v4_constraint,
@@ -491,12 +491,22 @@ def annotate_v4_with_constraint(genes_path, constraint_path):
     },
 )
 
+pipeline.add_task(
+    "annotate_grch38_genes_step_7",
+    annotate_gene_models_with_low_coverage_flag,
+    f"/{genes_subdir}/genes_grch38_annotated_7.ht",
+    {
+        "genes_path": pipeline.get_task("annotate_grch38_genes_step_6"),
+        "low_coverage_tsv_path": "gs://gnomad-v4-data-pipeline/inputs/v4.1.1/gnomad.v4.1.low_coverage_transcripts.tsv",
+    },
+)
+
 pipeline.add_task(
     "remove_grch38_genes_constraint_for_release",
     remove_gnomad_v4_constraint,
-    f"/{genes_subdir}/genes_grch38_annotate_5_removed_constraint",
+    f"/{genes_subdir}/genes_grch38_annotate_7_removed_constraint",
     {
-        "genes_path": pipeline.get_task("annotate_grch38_genes_step_5"),
+        "genes_path": pipeline.get_task("annotate_grch38_genes_step_7"),
     },
 )
 
@@ -563,7 +573,7 @@ def annotate_v4_with_constraint(genes_path, constraint_path):
 pipeline.set_outputs(
     {
         "genes_grch37": "annotate_grch37_genes_step_5",
-        "genes_grch38": "annotate_grch38_genes_step_6",
+        "genes_grch38": "annotate_grch38_genes_step_7",
         "base_transcripts_grch37": "extract_grch37_transcripts",
         "base_transcripts_grch38": "extract_grch38_transcripts",
         "transcripts_grch37": "annotate_grch37_transcripts",