bcbio / icr142-validation

Preparation scripts and bcbio integration for the ICR142 NGS validation series

ICR142 validation in bcbio

Support running an ICR142 validation using bcbio

http://f1000research.com/articles/5-386/v1

Running validation

This repository contains a full set of configuration files and BED/VCF validation sets to run an analysis with bcbio:

1. Obtain the ICR142 fastq files, which require applying for access. Move these to bcbiorun/input/fastqs

2. Run the analysis using an installed version of bcbio. This can run on a single machine using multiple cores or distributed on a cluster:

   cd bcbiorun/work
   bcbio_nextgen.py ../config/icr142.yaml -n 16
   

3. Summarize and plot the results:

   cd ../summarize
   bcbio_python ../../scripts/combine_samples.py
   bcbio_python ../../scripts/bcbio_validation_plot.py icr142-summary.csv
   

Results

Validation using bwa-mem and 3 variant callers (GATK HaplotypeCaller, FreeBayes and VarDict), including ensemble regions with calls in 2 of our 3 or 3 out of 3 callers. The majority of false positives are present in at least 2 callers, and many in all 3:

Truth set preparation

We prepared the truth set and analysis regions using the truth set calls from Supplemental table 1: scripts/icr_to_vcf.py created the VCF and BED files contained in the repository from the original table and a list of variants found to be homozygous (both in bcbiorun/input). The initial truth table does not have information about whether exepcted variants are homozygous or heterozygous so we ran an intial validation with everything heterozygous, then used scripts/find_hethomerrors.py to find those variants that are likely homozygous to reprepare the final truth set.