Regenie pipeline for LOF. Burden mode for regenie is used. This pipeline does all the work from annotation/vcfs to results.
The wdl(wdl/regenie_lof.wdl) runs the whole pipeline.
The first step is the extraction of LOF variants. The task extract_variants does the job. It also filters out common variants (regenie_lof.max_maf input) and low quality imputed variants (regenie_lof.extract_variants.info_filter)
chr10_1020846_AG_A IDI2 frameshift_variant
chr10_102733581_C_T SFXN2 stop_gained
chr10_103090970_C_CCAGA NT5C2 frameshift_variant
chr10_103246174_C_T RPEL1 stop_gained
chr10_103419670_C_T PDCD11 stop_gained
chr10_103449599_CAG_C CALHM2 frameshift_variant
chr10_103455672_G_A CALHM1 stop_gained
chr10_103473333_C_T CALHM3 stop_gained
chr10_113595655_G_A NRAP stop_gained
chr10_113597145_TG_T NRAP frameshift_variant
The variants are then passed to convert_vcf and then to merge to build a single bgen with only lof varirants.
From there a scatter takes each pheno and runs the regenie task, where the association takes place.
The task merge_results then produces all files needed for release:
- a readme file
- all hits sorted by logp
- merged logs
- a summary of the results only including sig hits and sorted by difference between mlgop and max(mlogp) of the variants in the gene
- the file required for sql import
Relevant inputs:
extract_variants.annot_file: File with lof, AF and info score data. All information is extracted automatically from header names.'
merge.bargs: the vcf--> bgen conversion params
extract_variants.lof_list: the list of lof keywords to include
regenie.cov_file: self explanatory, where the covariate and pheno data is located
regenie.masks_type: the bgen parameter that determines the merging logic for variants into genes
regenie.bargs : all other regenie args.