feat!: adding support for JSONL from v0.15 clinvar-this (#454)

build.rs

@@ -6,6 +6,11 @@ use std::{env, path::PathBuf};
 fn main() -> Result<(), anyhow::Error> {
     let root = PathBuf::from(env!("CARGO_MANIFEST_DIR")).join("protos");
     let proto_files = vec![
+        "annonars/clinvar_data/class_by_freq.proto",
+        "annonars/clinvar_data/clinvar_public.proto",
+        "annonars/clinvar_data/extracted_vars.proto",
+        "annonars/clinvar_data/gene_impact.proto",
+        "annonars/clinvar_data/phenotype_link.proto",
         "annonars/clinvar/minimal.proto",
         "annonars/clinvar/per_gene.proto",
         "annonars/clinvar/sv.proto",
@@ -51,7 +56,7 @@ fn main() -> Result<(), anyhow::Error> {
     let descriptor_set = std::fs::read(descriptor_path).unwrap();
     pbjson_build::Builder::new()
         .register_descriptors(&descriptor_set)?
-        .build(&[".annonars"])?;
+        .build(&[".annonars", ".clinvar_data"])?;
 
     Ok(())
 }

protos/annonars/clinvar/per_gene.proto

@@ -5,87 +5,24 @@ syntax = "proto3";
 package annonars.clinvar.per_gene;
 
 import "annonars/clinvar/minimal.proto";
-
-// Impact on gene in ClinVar.
-enum Impact {
-    // unknown
-    IMPACT_UNKNOWN = 0;
-    // 3' UTR variant
-    IMPACT_THREE_PRIME_UTR_VARIANT = 1;
-    // 5' UTR variant
-    IMPACT_FIVE_PRIME_UTR_VARIANT = 2;
-    // downstream gene variant
-    IMPACT_DOWNSTREAM_TRANSCRIPT_VARIANT = 3;
-    // frameshift variant
-    IMPACT_FRAMESHIFT_VARIANT = 4;
-    // inframe indel
-    IMPACT_INFRAME_INDEL = 5;
-    // start lost
-    IMPACT_START_LOST = 6;
-    // intron variant
-    IMPACT_INTRON_VARIANT = 7;
-    // missense variant
-    IMPACT_MISSENSE_VARIANT = 8;
-    // non-coding transcript variant
-    IMPACT_NON_CODING_TRANSCRIPT_VARIANT = 9;
-    // stop gained
-    IMPACT_STOP_GAINED = 10;
-    // no sequence alteration
-    IMPACT_NO_SEQUENCE_ALTERATION = 11;
-    // splice acceptor variant
-    IMPACT_SPLICE_ACCEPTOR_VARIANT = 12;
-    // splice donor variant
-    IMPACT_SPLICE_DONOR_VARIANT = 13;
-    // stop lost
-    IMPACT_STOP_LOST = 14;
-    // synonymous variant
-    IMPACT_SYNONYMOUS_VARIANT = 15;
-    // upstream gene variant
-    IMPACT_UPSTREAM_TRANSCRIPT_VARIANT = 16;
-}
-
-// Counts of variants by impact.
-message GeneImpactRecordCounts {
-    // Impact
-    Impact impact = 1;
-    // Counts for ACMG1..5
-    repeated uint32 counts = 2;
-}
-
-// Coarsened clinical significance
-enum CoarseClinicalSignificance {
-    // unknown
-    COARSE_CLINICAL_SIGNIFICANCE_UNKNOWN = 0;
-    // benign / likely benign
-    COARSE_CLINICAL_SIGNIFICANCE_BENIGN = 1;
-    // uncertain signifiance
-    COARSE_CLINICAL_SIGNIFICANCE_UNCERTAIN = 2;
-    // pathogenic / likely pathogenic
-    COARSE_CLINICAL_SIGNIFICANCE_PATHOGENIC = 3;
-}
-
-// Counts per frequency and impact.
-message GeneFreqRecordCounts {
-    // Frequency
-    CoarseClinicalSignificance coarse_clinsig = 1;
-    // Counts for ACMG1..5
-    repeated uint32 counts = 2;
-}
-
-// Variants for a given genome release / assembly.
-message GeneVariantsForRelease {
-    // Genome release / assembly
-    string genome_release = 1;
-    // Variants
-    repeated annonars.clinvar.minimal.Record variants = 2;
+import "annonars/clinvar_data/class_by_freq.proto";
+import "annonars/clinvar_data/extracted_vars.proto";
+import "annonars/clinvar_data/gene_impact.proto";
+
+// Extracted variants per release.
+message ExtractedVariantsPerRelease {
+    // Release version.
+    optional string release = 1;
+    // Variants per gene.
+    repeated clinvar_data.extracted_vars.ExtractedVcvRecord variants = 2;
 }
 
 // ClinVar detailed information per gene.
 message ClinvarPerGeneRecord {
     // Counts of variants per impact
-    repeated GeneImpactRecordCounts per_impact_counts = 1;
+    optional clinvar_data.gene_impact.GeneImpactCounts per_impact_counts = 1;
     // Counts of variants per impact / frequency
-    repeated GeneFreqRecordCounts per_freq_counts = 2;
+    optional clinvar_data.class_by_freq.GeneCoarseClinsigFrequencyCounts per_freq_counts = 2;
     // Variants for the given gene.
-    repeated GeneVariantsForRelease variants = 3;
+    repeated ExtractedVariantsPerRelease per_release_vars = 3;
 }

protos/annonars/clinvar_data/class_by_freq.proto

@@ -0,0 +1,29 @@
+// Protocol buffers for types for class-by-freq tool.
+
+syntax = "proto3";
+
+package annonars.clinvar_data.class_by_freq;
+
+// Enumeration for coarse-grain classification.
+enum CoarseClinicalSignificance {
+  // unspecified coarse clinical significance
+  COARSE_CLINICAL_SIGNIFICANCE_UNSPECIFIED = 0;
+  // Corresponds to "benign".
+  COARSE_CLINICAL_SIGNIFICANCE_BENIGN = 1;
+  // Corresponds to "uncertain".
+  COARSE_CLINICAL_SIGNIFICANCE_UNCERTAIN = 2;
+  // Corresponds to "pathogenic".
+  COARSE_CLINICAL_SIGNIFICANCE_PATHOGENIC = 3;
+}
+
+// Output record.
+message GeneCoarseClinsigFrequencyCounts {
+  // The gene HGNC ID.
+  string hgnc_id = 1;
+  // The counts for (likely) pathogenic.
+  repeated uint32 pathogenic_counts = 2;
+  // The counts for uncertain significance.
+  repeated uint32 uncertain_counts = 3;
+  // The counts for (likely) benign.
+  repeated uint32 benign_counts = 4;
+}