feat: build annonars regions (clingen dosage) (#67)

varfish-org · Nov 22, 2023 · 664295d · 664295d
1 parent 3fd72dd
commit 664295d
Show file tree

Hide file tree

Showing 11 changed files with 183 additions and 0 deletions.
diff --git a/Snakefile b/Snakefile
@@ -165,6 +165,11 @@ rule all:
         f"output/full/annonars/gnomad-sv-exomes-grch38-{DV.gnomad_cnv4}+{PV.annonars}/rocksdb/IDENTITY",
         f"output/full/annonars/gnomad-sv-genomes-grch37-{DV.gnomad_sv}+{PV.annonars}/rocksdb/IDENTITY",
         f"output/full/annonars/gnomad-sv-genomes-grch38-{DV.gnomad_sv4}+{PV.annonars}/rocksdb/IDENTITY",
+        # ----- sequence annotation
+        f"output/full/annonars/functional-grch37-{DV.refseq_fe_37}+{PV.annonars}/rocksdb/IDENTITY",
+        f"output/full/annonars/functional-grch38-{DV.refseq_fe_38}+{PV.annonars}/rocksdb/IDENTITY",
+        f"output/full/annonars/regions-grch37-{DV.today}+{PV.annonars}/rocksdb/IDENTITY",
+        f"output/full/annonars/regions-grch38-{DV.today}+{PV.annonars}/rocksdb/IDENTITY",
         # ----- conservation
         f"output/full/annonars/cons-grch37-{DV.ucsc_cons_37}+{PV.annonars}/rocksdb/IDENTITY",
         f"output/full/annonars/cons-grch38-{DV.ucsc_cons_38}+{PV.annonars}/rocksdb/IDENTITY",
@@ -386,6 +391,8 @@ include: "rules/output/annonars/gnomad_mtdna.smk"
 include: "rules/output/annonars/gnomad_sv.smk"
 include: "rules/output/annonars/helix.smk"
 include: "rules/output/annonars/genes.smk"
+include: "rules/output/annonars/functional.smk"
+include: "rules/output/annonars/regions.smk"
 # ---- worker
 include: "rules/output/worker/patho_mms.smk"
 include: "rules/output/worker/clinvar.smk"

diff --git a/download_urls.yml b/download_urls.yml
@@ -1,3 +1,12 @@
+- url: https://ftp.ncbi.nlm.nih.gov/genomes/all/annotation_releases/9606/105.20201022/GCF_000001405.25_GRCh37.p13/GCF_000001405.25_GRCh37.p13_genomic.gff.gz
+  excerpt_strategy:
+    strategy: gz-head
+    count: 1000
+- url: https://ftp.ncbi.nlm.nih.gov/genomes/all/annotation_releases/9606/110/GCF_000001405.40_GRCh38.p14/GCF_000001405.40_GRCh38.p14_genomic.gff.gz
+  excerpt_strategy:
+    strategy: gz-head
+    count: 1000
+
 - url: https://storage.googleapis.com/gcp-public-data--gnomad/legacy/exac_browser/ExAC.r1.sites.vep.vcf.gz
   excerpt_strategy:
     strategy: gz-head

diff --git a/excerpt-data/98935d27cc8f0dc0/GCF_000001405.40_GRCh38.p14_genomic.gff.gz b/excerpt-data/98935d27cc8f0dc0/GCF_000001405.40_GRCh38.p14_genomic.gff.gz
diff --git a/excerpt-data/98935d27cc8f0dc0/url.txt b/excerpt-data/98935d27cc8f0dc0/url.txt
diff --git a/excerpt-data/f0ed4b0862f1b46b/GCF_000001405.25_GRCh37.p13_genomic.gff.gz b/excerpt-data/f0ed4b0862f1b46b/GCF_000001405.25_GRCh37.p13_genomic.gff.gz
diff --git a/excerpt-data/f0ed4b0862f1b46b/url.txt b/excerpt-data/f0ed4b0862f1b46b/url.txt
diff --git a/rules/output/annonars/functional.smk b/rules/output/annonars/functional.smk
@@ -0,0 +1,67 @@
+## Rules to create build annonars functional annotation database..
+
+
+rule work_annonars_functional_download_37:  # -- download functional data for GRCh37
+    output:
+        "work/download/refseq/grch37/{version}/{assembly}_genomic.gff.gz",
+    shell:
+        r"""
+        wget -O {output} \
+            https://ftp.ncbi.nlm.nih.gov/genomes/all/annotation_releases/9606/{wildcards.version}/{wildcards.assembly}/{wildcards.assembly}_genomic.gff.gz
+        """
+
+
+rule work_annonars_functional_download_38:  # -- download functional data for GRCh37
+    output:
+        "work/download/refseq/grch38/{version}/{assembly}_genomic.gff.gz",
+    shell:
+        r"""
+        wget -O {output} \
+            https://ftp.ncbi.nlm.nih.gov/genomes/all/annotation_releases/9606/{wildcards.version}/{wildcards.assembly}/{wildcards.assembly}_genomic.gff.gz
+        """
+
+
+def output_annonars_functional_input(wildcards):
+    if wildcards.genome_release == "grch37":
+        return f"work/download/refseq/grch37/{DV.refseq_fe_37}/GCF_000001405.25_GRCh37.p13_genomic.gff.gz"
+    else:
+        return f"work/download/refseq/grch38/{DV.refseq_fe_38}/GCF_000001405.40_GRCh38.p14_genomic.gff.gz"
+
+
+rule output_annonars_functional:  # -- build annonars functional RocksDB file
+    input:
+        output_annonars_functional_input,
+    output:
+        rocksdb_identity=(
+            "output/full/annonars/functional-{genome_release}-{v_refseq}+{v_annonars}/"
+            "rocksdb/IDENTITY"
+        ),
+        spec_yaml=(
+            "output/full/annonars/functional-{genome_release}-{v_refseq}+{v_annonars}/spec.yaml"
+        ),
+    wildcard_constraints:
+        v_refseq=RE_VERSION,
+        v_annonars=RE_VERSION,
+    shell:
+        r"""
+        export TMPDIR=$(mktemp -d)
+        trap "rm -rf $TMPDIR" EXIT
+
+        zgrep '^#\|RefSeqFE' {input} > $TMPDIR/tmp.gff
+
+        annonars functional import -vvv \
+            --genome-release {wildcards.genome_release} \
+            --path-in-gff $TMPDIR/tmp.gff \
+            --path-out-rocksdb $(dirname {output.rocksdb_identity})
+
+        varfish-db-downloader tpl \
+            --template rules/output/annonars/functional.spec.yaml \
+            --value today={TODAY} \
+            \
+            --value version={wildcards.v_refseq}+{wildcards.v_annonars} \
+            --value v_refseq={wildcards.v_refseq} \
+            \
+            --value v_annonars={wildcards.v_annonars} \
+            --value v_downloader={PV.downloader} \
+        > {output.spec_yaml}
+        """
diff --git a/rules/output/annonars/functional.spec.yaml b/rules/output/annonars/functional.spec.yaml
@@ -0,0 +1,16 @@
+dc.identifier: annonars/functional:{{ version }}-{{ genome_release }}
+dc.title: annonars functional elements RocksDB
+dc.creator: VarFish Developer Teams
+dc.format: application/x-rocksdb
+dc.date: {{ today }}
+x-version: {{ version }}
+x-genome-release: {{ genome_release }}
+dc.description: |
+  RocksDB built from RefSeq Functional Elements (and other sources in
+  the future).
+dc.source:
+  - PMID:34876495
+  - https://www.ncbi.nlm.nih.gov/refseq/
+x-created-from:
+  - name: RefSeq Functional Elements
+    version: {{ v_refseq }}
diff --git a/rules/output/annonars/regions.smk b/rules/output/annonars/regions.smk
@@ -0,0 +1,52 @@
+## Rules to create build annonars regions annotation database..
+
+
+rule work_annonars_regions_download:  # -- download clingen regions
+    output:
+        "work/download/clingen/{genome_release}/{today}/ClinGen_region_curation_list_{genome_release}.tsv",
+    shell:
+        r"""
+        if [[ "{wildcards.genome_release}" == "grch38" ]]; then
+            GENOME=GRCh37
+        else
+            GENOME=GRCh38
+        fi
+
+        wget -O {output} \
+            ftp://ftp.clinicalgenome.org/ClinGen_region_curation_list_$GENOME.tsv
+        """
+
+
+rule output_annonars_regions:  # -- build annonars regions RocksDB file
+    input:
+        "work/download/clingen/{genome_release}/{date}/ClinGen_region_curation_list_{genome_release}.tsv",
+    output:
+        rocksdb_identity=(
+            "output/full/annonars/regions-{genome_release}-{date}+{v_annonars}/" "rocksdb/IDENTITY"
+        ),
+        spec_yaml=("output/full/annonars/regions-{genome_release}-{date}+{v_annonars}/spec.yaml"),
+    wildcard_constraints:
+        v_refseq=RE_VERSION,
+        v_annonars=RE_VERSION,
+    shell:
+        r"""
+        if [[ "$(date +%Y%m%d)" != "{wildcards.date}" ]] && [[ "{FORCE_TODAY}" != "True" ]]; then
+            >&2 echo "{wildcards.date} is not today"
+            exit 1
+        fi
+
+        annonars regions import -vvv \
+            --genome-release {wildcards.genome_release} \
+            --path-in-clingen {input} \
+            --path-out-rocksdb $(dirname {output.rocksdb_identity})
+
+        varfish-db-downloader tpl \
+            --template rules/output/annonars/regions.spec.yaml \
+            --value today={TODAY} \
+            \
+            --value version={wildcards.date}+{wildcards.v_annonars} \
+            \
+            --value v_annonars={wildcards.v_annonars} \
+            --value v_downloader={PV.downloader} \
+        > {output.spec_yaml}
+        """
diff --git a/rules/output/annonars/regions.spec.yaml b/rules/output/annonars/regions.spec.yaml
@@ -0,0 +1,14 @@
+dc.identifier: annonars/regions:{{ version }}-{{ genome_release }}
+dc.title: annonars regions annotation RocksDB
+dc.creator: VarFish Developer Teams
+dc.format: application/x-rocksdb
+dc.date: {{ today }}
+x-version: {{ version }}
+x-genome-release: {{ genome_release }}
+dc.description: |
+  RocksDB with region annotation.
+dc.source:
+  - https://search.clinicalgenome.org/kb/gene-dosage
+x-created-from:
+  - name: ClinGen Region Dosage Pathogenicity
+    version: {{ today }}
diff --git a/varfish_db_downloader/versions.py b/varfish_db_downloader/versions.py
@@ -111,6 +111,10 @@ class DataVersions:
     #: Marker file for the tracks version.  This allows us to update the
     #: tracks BED files later on.
     tracks: str
+    #: RefSeq functional elements for GRCh37.
+    refseq_fe_37: str
+    #: RefSeq functional elements for GRCh38.
+    refseq_fe_38: str
 
 
 #: The data versions to use.
@@ -158,6 +162,8 @@ class DataVersions:
     clinvar_release=CLINVAR_RELEASE,
     clinvar_version=CLINVAR_VERSION,
     tracks="0",
+    refseq_fe_37="105.20201022",
+    refseq_fe_38="110",
 )