Kipoi-GWAS is a snakemake pipeline which

• downloads UK BioBank association files

• merges the annotation files with variant effect predictions using Kipoi

• runs FGWAS to fine-map variants

• generates a report

A sample report can be found here: src/fgwas_plot.ipynb .

General workflow of Kipoi-GWAS

Installation

1. Clone the git repo.

git clone https://github.com/NCBI-Hackathons/Kipoi-GWAS

1. Install conda environment.

conda env create -f environment.yml

1. Install the python package

cd Kipoi-GWAS; pip install .

Folder structure

Here is the complete folder structure.

input/
 UKBB/ # UK-biobank phenotypes downloaded from the UK Biobank site. Select phenotypes of interest. TODO
 {phenotype}.gwas.imputed_v3.both_sexes.tsv.bgz	{phenotype}.gwas.imputed_v3.both_sexes.tsv # after untarring the .bgz file
 anno/ # variant annotation of the phenotypes found in UKBB
 kipoi/ # annotated variants using `kipoi veff score_variants`
 subset/
 {chr}/
 {model}.tsv.gz
output/
 {phenotype}/
 subset/
 {chr}/
 {run-id}/
 metadata.json # information about the run
 fgwas/
 input/ # input tables for fgwas
 output/ # output of fgwas, includes default outputs
 report/ # reports
 fig1.ipynb
 fig1.html

Placeholders

• {phenotype} - the UKBB phenotype code goes here

• {chr} - results are generated for associations in a particular chromosome. Example: chr12

• {run-id} - this is a manually selected placeholder which allows for the pipeline to be run for different combinations of variant annotations