Imputation Workflow

This repo contains the workflow which was developed as part of the H3ABioNet Hackathon held in Pretoria, SA in 2016.

We track our open tasks using github's issues
The 1000ft view is located on our Trello board.

Setup (native cluster)

Headnode

Nextflow (can be installed as local user)
NXF_HOME needs to be set, and must be in the PATH
Note that we've experienced problems running Nextflow when NXF_HOME is on an NFS mount.
The Nextflow script also needs to be invoked in a non-NFS folder
Java 1.8+

Compute nodes

The compute nodes need access to shared storage for input, references, output
The following commands need to be available in PATH on the compute nodes
- impute2 from IMPUTE2
- plink from PLINK 1.9+
- vcftools from VCFtools
- bcftoolsfrom bcftools
- bgzip from htslib
- eagle from Eagle
- python2.7

Getting started

Basic test

Clone this repo
Run the "tiny" dataset included

nextflow run imputation.nf -c nextflow.test.tiny.config

check for results in outfolder

wc -l output/impute_results/FINAL_VCFS/*

Larger dataset

Download this slightly larger dataset: small.tar.bz2 and extract into the samples folder
Run this "small" dataset with

nextflow run imputation.nf -c nextflow.test.small.config

check for results in outfolder

wc -l output/impute_results/FINAL_VCFS/*

AWS Batch

This workflow can be run on AWS Batch, an Amazon service that streamlines the running of containerised workflows. Nextflow

Set up the AWS environment. The requirements are:

an AWS account with credits. See the AWS grants page for grants
a configured Compute Environment and associated AWS Job Queue
a configured S3 bucket

Populate the awsbatch profile in

aws.region = 'eu-west-1'
aws.client.storageEncryption = 'AES256'
process.queue = 'large'
executor.name = 'awsbatch'
executor.awscli = '/home/ec2-user/miniconda/bin/aws'

run with the awsbatch profile

nextflow run imputation.nf -c nextflow.test.small.config -profile awsbatch

nknowlton / chipimputation