Variant interpretation

This repository contains utilities for consolidating, annotating and filtering SNVs/indels and structural variants (SVs). This project is a work in progress.

Table of Contents

• Requirements
• Family filtering
• De novo SV discovery
• Contact and credits

Requirements

Deployment and execution:

• A Google Cloud account.
• A workflow execution system supporting the Workflow Description Language (WDL), such as Cromwell (v36 or higher). A dedicated server is highly recommended.
• cromshell for interacting with a dedicated Cromwell server or a Terra account
• A vcf file output from the GATK-SV pipeline

Annotation and Filtering of Possible Clinically Relevant SVs

Additional annotations and filtering are performed for each family given a input pedigree file. The input is a joint-genotyped VCF obtained from GATK-SV and the output is a reformated BED file.

Filtering

• AF <= 0.3 and gnomAD_V2_AF <= 0.01 OR SV overlaps a genomic disorder region (explained below).

Annotation

• Genomic disorders: all SVs overlapping a list of genomic disorders regions are flagged in the final output. Currenly it requires a minimum reciprocarl overlap with the GD of 30%.
• Constraint: extra columns are added for SVs involving genes that are constrained for LoF or missense variantsion (using pLI, pRec and EO metrics).
• Gene list: a colum that determines if each SV involves a gene present in the gene list provided is included (Note: this is currently using HGNC Symbols).
• OMIM: a colum that determines if each SV involves a gene present in OMIM is included (Note: this is currently using HGNC Symbols).
• Mode of Inheritance:
  • Absent in unaffected: a variant is flagged if it's present in unaffected individuals, and present only in affected. This is particularly relevant for trios and other family structures to identify de novo variants or inherited from affected parents (Column name: FILT_ABSENT_UNAFF)
  • Multiple SVs in same gene: a variant is flagged if there are multiple variants in the same gene for the same individual. If trio is available, it only returns if they are in trans (Column name: MULT_HIT)
  • Autosomal recessive: a variant is flagged if genotype for affected individual/s is 1/1 and is not for unaffected (Column name: FILT_AR)
  • X-linked: a variant is flagged as FILT_XLR column if they are in male affected individual/s and they are 1/1 or 1 genotypes.
  • Inherited: possible intereting inherited variants from unaffected parents are flagged in the FILT_INHERITED column if they are in the input gene list AND the cohort allele count <= 10 AND the number of unaffected indivuals in the cohort is <= 5.

De novo SV discovery

Filters de novo SV obtained from GATK-SV to improve specificity of de novo SV calls. The filtering is based on a combination of site-specific and sample-specific metrics:

• Frequency filter
  • gnomAD AF <= 0.01 OR in genomic disorder (GD)
  • Other parents in the cohort: AF <= 0.01 OR AC <= 3
• Absent call in both parents
• By type specific filters:
  • CPX,CTX,INV: keep all
  • INS: require raw evidence support
  • DEL,DUP: require raw evidence support AND,
    • Large (>5KB) CNVs require depth support AND no parent overlap (50%) AND no RDCN evidence in parent
    • Small (<1Kb) CNVs with SR-only evidence require input on both sides

Input data:

• VCF file: annotated with GATK-SV AnnotateVcf.
• PED file: with columns FamID, IndividualID, FatherID, MotherID, Gender, Affected
• Disorder input: bed file of SV sites that are genomic disorder regions
• Directory with raw calls from GATK-SV GatherBatchEvidence module with the following columns: chrom, start, end, svtype
• Raw files: txt file with first column as the batches and second column as the raw files generated from [GATK-SV ClusterBatch] (https://github.com/broadinstitute/gatk-sv/blob/main/wdl/ClusterBatch.wdl). This list must contain the raw files from all callers except for depth
• Depth raw files: txt file with first column as the batches and second column as the depth raw files generated from GATK-SV ClusterBatch. This list must contain the raw files from only the depth caller
• Blacklist regions: bed file with a list of regions you want to exclude from the de novo analysis with the following columns: chrom, start, end
• Coverage matrices: txt file with first column as batch name, the second column as coverage matrix generated from GATK-SV GatherBatchEvidence, and third column as coverage matrix index file generated from GATK-SV GatherBatchEvidence
• Sample and batch input: txt file with samples in first column and their repsective batches in second column
• Python config file: a python file defining input parameters

Execution

The main scripts to run this analysis is runDeNovoSVs.wdl, which can be run on Terra or with Cromshell with the following commands:

> git clone https://github.com/talkowski-lab/variant-interpretation.git
> cd variant-interpretation/wdl
> zip dependencies.zip *
> cromshell submit runDeNovoSVs.wdl /path/to/denovo-svs.json /path/to/config.json dependencies.zip

Contact and credits

Copyright (c) 2022 Talkowski Lab and The Broad Institute of M.I.T. and Harvard
Contact: Alba Sanchis-Juan

Variant interpretation team: Alba Sanchis-Juan, Nehir Kurtas, Eren Shin, Arthur Lee, Emily Liau, Nicole Calamari, Harrison Brand.