VarFish Annotator

Annotation of VCF file for import into VarFish (through Web UI).

Supported Databases

ExAC r1.0
gnomAD exomes r2.1
gnomAD genomes r2.1
clinvar
thousand genomes
hgmd_public from ENSEMBL r75

Supported Variant Callers

VarFish annotator uses HTSJDK for reading variant call format (VCF) files. HTSJDK supports reading VCF v4.3 so the output of any tool that produces well-formed VCF can be read. VCF itself only specifies relatively few required fields and callers may use fields in a slightly different way. We thus document below what fields are used/interpreted by VarFish Annotator to prepare the files for VarFish.

Small Variants

The following fields are considered:

CHROM Will be written out in a normalized way, depending on the genome build. That the chr prefix will be presenet for GRCh38 and it will be absent for GRCh37.
POS The 1-based chromosomal position is written out.
REF The reference allele will be written out.
ALT For each alternative allele and each gene, a record is written out. Asterisk alleles will be ignored (see Broad's GATK documentation on this).
FORMAT and per SAMPLE
- GT The genotype is written out with phasing information.
- AD The allelic depth of the alternative allele of the current record is written out.
- DP The total depth is written out to compute the allelic balance of the alternative allele of the current record is written out.
- GQ The genotype quality is written out.

Structural Variants / Copy Number Variants

Supported Callers and Caller Annotation

The following variant callers are explicitely supported.

Delly 2 (SVs)
Dragen CNV caller
Dragen SV caller
Manta
GATK gCNV
XHMM (deprecated)

In the other cases, VarFish annotator will fall back to a "generic" import where only the per-sample fields GT, FT, and GQ are interpreted. Your caller should also write out INFO/END, INFO/SVTYPE, and INFO/SVLEN as defined by VCF4.2

VarFish Annotator will look at the field INFO/SVMETHOD to annotate calls with the caller where the call originated from. If this field is empty then you should define --default-sv-method so you get appropriately labeled output. If you have any problem with your data then please tell us by opening a GitHub issue.

Interpretation of top-level and INFO VCF fields

The following fields are considered:

CHROM Will be written out in a normalized way, depending on the genome build. That the chr prefix will be presenet for GRCh38 and it will be absent for GRCh37.
POS The 1-based chromosomal position is written out.
REF The reference allele will be written out.
ALT For each alternative allele and each gene, a record is written out. Asterisk alleles will be ignored (see Broad's GATK documentation on this).
INFO/CHR2 Second chromosome of the SV, if not on the same chromosome.
INFO/END End position of the SV.
INFO/CT Paired-endsignature induced connection type.
INFO/CIPOS Confidence interval around the start point of the SV.
INFO/CIEND Confidence interval around the end point of the SV.
INFO/SVMETHOD The name of the caller that was used.

Interpretation of FORMAT and per sample fields

Common
- GT Genotype, written as gt
- FT Per-genotype filter values, written as ft
- GQ Phred-scaled genotype quality, written as gq
Delly2
- DR Reference pairs, written as pec = DR + DV
- DV Variant pairs, written as pev
- RR Reference junction count, written as src = RR + RV
- RV Variant junction count, written as srv
- RDCN Copy number estimate, written as cn
Dragen CNV
- SM Average normalized overage, written as anc
- BC Bucket count, written as point count pc
- PE Discordante read count at start/end, written as pev = PE[0] + PE[1]
Dragen SV
- PR Paired read of reference and variant, written as pec = PR[0] + PR[1] and pev = PR[1]
- SR Paired read of reference and variant, written as src = SR[0] + SR[1] and srv = SR[1]
For GATK gCNV
- CN Integer copy number, written as cn
- NP Number of points in segment, written as np
Manta (equivalent to Dragen SV)
For XHMM
- RD Average normalized coveage, written as an

Example

The following will create varfish-annotator-db-1906.h2.db and fill it.

# DOWNLOAD=path/to/varfish-db-downloader
# ANNOTATOR_VERSION=0.9
# ANNOTATOR_DATA_RELEASE=1907
# java -jar varfish-annotator-cli-$ANNOTATOR_VERSION-SNAPSHOT.jar \
      init-db \
      --db-release-info "varfish-annotator:v$ANNOTATOR_VERSION" \
      --db-release-info "varfish-annotator-db:r$ANNOTATOR_DATA_RELEASE" \
      \
      --ref-path /fast/projects/cubit/18.12/static_data/reference/GRCh37/hs37d5/hs37d5.fa \
      \
      --db-release-info "clinvar:2019-02-20" \
      --clinvar-path $DOWNLOAD/GRCh37/clinvar/latest/clinvar_tsv_main/output/clinvar_allele_trait_pairs.single.b37.tsv \
      --clinvar-path $DOWNLOAD/GRCh37/clinvar/latest/clinvar_tsv_main/output/clinvar_allele_trait_pairs.multi.b37.tsv \
      \
      --db-path ./varfish-annotator-db-$ANNOTATOR_DATA_RELEASE \
      \
      --db-release-info "exac:r1.0" \
      --exac-path $DOWNLOAD/GRCh37/ExAC/r1/download/ExAC.r1.sites.vep.vcf.gz \
      \
      --db-release-info "gnomad_exomes:r2.1" \
      $(for path in $DOWNLOAD/GRCh37/gnomAD_exomes/r2.1/download/gnomad.exomes.r2.1.sites.chr*.normalized.vcf.bgz; do \
          echo --gnomad-exomes-path $path; \
      done) \
      \
      --db-release-info "gnomad_genomes:r2.1" \
      $(for path in $DOWNLOAD/GRCh37/gnomAD_genomes/r2.1/download/gnomad.genomes.r2.1.sites.chr*.normalized.vcf.bgz; do \
          echo --gnomad-genomes-path $path; \
      done) \
      \
      --db-release-info "thousand_genomes:v3.20101123"
      $(for path in $DOWNLOAD/GRCh37/thousand_genomes/phase3/ALL.chr*.phase3_shapeit2_mvncall_integrated_v5a.20130502.sites.vcf.gz; do \
          echo --thousand-genomes-path $path; \
      done) \
      \
      --db-release-info "hgmd_public:ensembl_r75" \
      --hgmd-public $DOWNLOAD/GRCh37/hgmd_public/ensembl_r75/HgmdPublicLocus.tsv

Formatting Source Code

# mvn com.coveo:fmt-maven-plugin:format -Dverbose=true

Tests

The folder /tests contains some data sets that are appropriate for system (aka "end-to-end") tests of the software.

hg19-chr22 -- This folder contains examples for annotating GATK HC and Delly2 calls on the first 20MB of chr22. Only the variants overlapping with ADA2 and GAB4 are used.

You can build the data sets with the build.sh script that is available in each folder. This script also serves for documenting the test data's provenance. The Jannovar software must be available as jannovar (e.g., through bioconda) on your PATH and you will need samtools.

Using JDK >=18

The tests use junit5-system-exit for detecting System.exit() calls. In JDK 18 you have to use the -Djava.security.manager=allow flag. Issue tginsberg/junit5-system-exit#10 tracks this issue.

Developing on Windows

There is an issue with removing temporary directories on Windows. Apparently, HTSJDK does not properly close files. Set -Djunit.jupiter.tempdir.cleanup.mode.default=NEVER to work around this issue.

varfish-org / varfish-annotator

readme