Skip to content

Latest commit

 

History

History
57 lines (37 loc) · 3 KB

README.md

File metadata and controls

57 lines (37 loc) · 3 KB
Raw

Cohort SV detection pipeline

Table of contents

  1. Overview
  2. WDL scripts
  3. Docker images

Overview

This repository contains pipeline scripts for structural variation detection in large cohorts. The pipeline is designed for Illumina paired-end whole genome sequencing data, preferably with at least 30x sequence coverage. Data inputs should be a set of sorted CRAM files, aligned with BWA-MEM.

This pipeline detects structural variation based on breakpoint sequence evidence using the LUMPY algorithm. Structural variant (SV) breakpoints are then unified and merged using the SVTools workflow, followed by re-genotyping with SVTyper and read-depth annotation with CNVnator. Finally, SV types are reclassified based on the concordance between read-depth and breakpoint genotype.

Additional details on the SVTools pipeline are available in the SVTools tutorial.

Workflow

WDL scripts

Pipeline scripts (in WDL format) are available in the scripts directory. These scripts can be launched using Cromwell (version 25 or later).

While the SV pipeline can be run in its entirety via the SV_Pipeline_Full.wdl script, we recommend running the pipeline in three stages to enable intermediate quality control checkpoints.

1. Pre_Merge_SV.wdl

For each sample:

  • Generate LUMPY inputs from an aligned and sorted CRAM file
  • SV discovery with LUMPY
  • Preliminary SV genotyping with SVTyper
  • Generate CNVnator histogram files

After this step, we recommend performing quality control checks on each sample before merging them into the cohort-level VCF (step 2).

2. Merge_SV.wdl

This step merges the sample-level VCF files from step 1 using the LUMPY breakpoint probability curves to produce a single cohort-level VCF.

3. Post_Merge_SV.wdl

This step re-genotypes each sample at the sites in the cohort-level VCF file from step 2, and then combines the results into the final VCF.

For each sample:

  • Re-genotype each SV using SVTyper
  • Annotate the read-depth at each SV using CNVnator
  • Generate a .ped file of sample names and sexes

For the cohort:

  • Combine the re-genotyped VCFs into a single cohort-level VCF
  • Prune overlapping SVs
  • Classify SV type based on the concordance between variant genotypes and read-depths
  • Sort and index the VCF

Docker images

  • Docker images for this pipeline are available at https://hub.docker.com/u/halllab.
  • Dockerfiles for these containers are available in the docker directory.
  • WDL test scripts for each of these Docker containers are available in the test directory.