Skip to content

Latest commit

 

History

History
96 lines (64 loc) · 3.53 KB

README.md

File metadata and controls

96 lines (64 loc) · 3.53 KB
Raw

lopass

Pipeline for low-coverage genotyping, imputation, phasing.

Step 1: gVCF Variant Calling (GATK4)

lopass-gatk.py A gVCF file is created calling variation within a BAM file.

Step 2: Extraction of genotype likelihoods (PLs) for panel variants

lopass-genotype.py A VCF file is created with PLs for all variants in a reference panel.

Given a reference panel (normalized indels, multi-allelics decomposed) and a single-sample gVCF file genotypes the positions in the reference panel given the evidence in the gVCF file. The gVCF file should not be decomposed.

Step 3: Imputation and Phasing

lopass-glimpse.py Imputation and phasing is performed on the resulting VCF file.

Installation

Lopass requires a Linux system with configured to compile htslib and boost, Python3 and a number of third-party optionally managed by conda.

Dependencies

  • conda (system installation)
  • boost (bundled)
  • htslib (bundled)
  • Python3 (managed via conda)
  • cyvcf2 (managed via conda)
  • gatk4 (managed via conda)

Building

  1. Install conda
  2. clone this github repository (recursively to get the correct GLIMPSE version) git clone https://github.com/mctp/lopass --recursive
  3. Execute build.sh script, this will compile GLIMPSE
  4. Execute setup-conda.sh this will create a lopass environment. Alternatively, install Python3, cyvcf2 and GATK4 manually.

Running

Before running lopass setup paths and activate the conda environment:

source env.sh

Lopass is currently only compatible with GRCh38-style chromosome names we need to re-code the genetic maps included in GLIMPSE

scripts/glimpse-recode-maps.sh GLIMPSE/maps/genetic_maps.b38 data/hg38/maps

Imputation panels need to be chunked by GLIMPSE prior to running:

scripts/glimpse-chunk-panel.sh <panel> data/hg38/chunks

Details

Panel variants are annotated with genotype calls (GT) and likelihoods (PL) from sample gVCF file. A panel variant may overlap a gVCF reference block or variant site.

  1. Panel variant overlaps a reference block. Every panel variant within a reference block will have a homozygous reference genotype and the same PL based on the catch-all <NON_REF> ALT.
  2. Panel variant matches a sample variant. Two variants are defined as matching if and only if:
  • they match on POS and REF
  • The panel ALT is the most-likely ALT in the sample variant
  • The sample genotype is homozygous reference, heterozygous, or homozygous alternative for the panel ALT

In all other cases, the site is returned as unobserved.

Sex chromosomes

Chromosome X is fully supported with 'correct' handling of XX and XY individuals and PAR1 and PAR2 regions.

Chromosome Y is not supported.

Example

Step 1:

./lopass-gatk.py -t1 -p4 -L chr22:1-18709565:2:2 -R grch38_full_analysis_set_plus_decoy_hla.fa --sex XX -i data/chr22-tiny/NA12878-10M-chr22-tiny.bam -o data/chr22-tiny/NA12878-10M-chr22-tiny

Step 2:

./lopass-genotype.py data/chr22-tiny/kg-hg38-nygc2020-chr22-tiny-sites.bcf data/chr22-tiny/NA12878-10M-chr22-tiny.vcf.gz -r chr22:1-18709565 | bcftools view -Oz -o data/chr22-tiny/NA12878-10M-chr22-tiny-call.vcf.gz
tabix data/chr22-tiny/NA12878-10M-chr22-tiny-call.vcf.gz

Step 3: Make sure to run the above preping step of generating maps and chunked panels prior to this step.

./lopass-glimpse.py -c data/chr22-tiny/chunks -m data/chr22-tiny/maps data/chr22-tiny/kg-hg38-nygc2020-chr22-tiny-genotypes-noNA12878.bcf  data/chr22-tiny/NA12878-10M-chr22-tiny-call.vcf.gz -o data/chr22-tiny/NA12878-10M-chr22-tiny-glimpse -r chr22 -d