Variant annotation

Simple variant annotation.

• Variants are annotated with their predicted impact on the genes using SNPeff
• Then with public variant databases:
  • Small variants are annotated with with gnomAD's allele frequencies, conservation/CADD/MetaRNN scores, and ClinVar's clinical significance.
  • Structural variants are annotated with the frequency in SV catalogs, and their overlap with dbVar clinical SVs, or DGV.
• Structural variants can also be re-genotyped from the long reads to compute the read support and help rank/filter them by confidence.

Inputs

Gene annotation

If SNPEFF_DB and SNPEFF_DB_NAME are provided, SNPeff will annotate variants with their impact based on the gene annotation.

Small variant databases

If GNOMAD_VCF, GNOMAD_VCF_INDEX, CLINVAR_VCF, CLINVAR_VCF_INDEX, DBNSFP_DB, and DBNSFP_DB_INDEX are provided, SNPeff/SNPsift will annotate small variants with:

• their frequency in gnomAD
• their presence and clinical significance in ClinVar
• the predicted impact based on conservation (GERP++), CADD, or MetaRNN.

Structural variant databases

If SV_DB_RDATA is provided, SVs are annotated with:

• the frequency of similar SVs in public SV catalogs
• their similarity with variants in DGV
• their similarity with the Clinical SV dataset from dbVar

Structural variant validation

If BAM, BAM_INDEX, and REFERENCE_FASTA are provided, SVs will be re-genotyped from the long reads using local pangenomes built with vg. Two new INFO fields will represent the read support for the SV allele:

• RS_PROP with the proportion of supporting reads.
• RS_AD with the read support for the reference and alternate alleles (e.g. 3,5 for 3 reference-supporting reads and 5 SV-supporting reads).

Prepare annotation files

gnomAD

To reduce the size of the VCF with the variants in gnomAD, keeping only the field that we want to use (AF):

for CHR in `seq 1 22` X Y
do
    curl https://storage.googleapis.com/gcp-public-data--gnomad/release/3.0/vcf/genomes/gnomad.genomes.r3.0.sites.chr${CHR}.vcf.bgz | zcat | bcftools annotate -x ^INFO/AF,ID,FILTER,QUAL -e 'FILTER~"AC0"' -Oz -o gnomad_chr.genomes.r3.0.sites.chr${CHR}.vcf.gz
done
bcftools concat gnomad_chr.genomes.r3.0.sites.chr*.vcf.gz | bcftools sort -Oz -o gnomad.genomes.r3.0.sites.small.vcf.bgz -m 3G
rm gnomad_chr.genomes.r3.0.sites.chr*.vcf.gz

The final bgzipped VCF is only 4.5G.

dbNSFP

To reduce the size of the dbNSFP database used by SNPeff:

## download dbNSFP zip from https://sites.google.com/site/jpopgen/dbNSFP
## list files in the zip
unzip -l dbNSFP4.4a.zip
## list columns in full txt file. helps finding/double-checking the column numbers we want to keep
unzip -p dbNSFP4.4a.zip dbNSFP4.4a_variant.chr1.gz | zcat | head -1 | awk 'BEGIN{RS="\t"}{N=N+1;print N" "$0}' | less
## keep only some columns. gzip to control the size of temporary files and use multi-threaded gzip (pigz) to speed up a bit
unzip -p dbNSFP4.4a.zip dbNSFP4.4a_variant.chr1.gz | zcat | head -1 | cut -f 1-4,77-79,128-130,165-167,676 | gzip > temp.txt.gz
for CHR in `seq 1 22` X Y M
do
    unzip -p dbNSFP4.4a.zip dbNSFP4.4a_variant.chr${CHR}.gz | zcat | sed 1d | cut -f 1-4,77-79,128-130,165-167,676 | pigz -c -p 4 >> temp.txt.gz
done
## bgzip instead of gzip
unpigz -c -p 4 temp.txt.gz | bgzip > dbNSFP4.4a.small.txt.gz
## index with tabix
tabix -b 2 -e 2 -s 1 dbNSFP4.4a.small.txt.gz

If we keep only information about GERP, CADD, MetaRNN, and ALFA, the file size decreases from 35G to about 2G.

Test locally

## download GRCh38.105 database
wget https://snpeff.blob.core.windows.net/databases/v5_1/snpEff_v5_1_GRCh38.105.zip

miniwdl run --as-me -i test/test.inputs.json wdl/workflow.wdl