Introduction

This is a Nextflow pipeline to run Genome Wide Association Studies (GWAS) in Arabidopsis. It uses the Limix framework for linear mixed models and supports both single-trait and multi-trait analysis while correcting for population structure using an IBS (identity-by-state) kinship matrix.

Summary

Installation

The pipeline is bundled with a container image containing all necessary dependencies. Therefore it is required to have Nextflow installed, along with one of the following Container engines: Docker, Podman, Singularity or Charliecloud.

To get the pipeline up an running, first clone the repository:

git clone [email protected]:beckerlab/gwas-nf.git

Note: The repository contains genotype information from the 1001genomes project (The 1001 Genomes Consortium, 2016), as well as the Regmap 250K panel (Horton et al., 2012). They are stored as hdf5 files under assets/genotypes. If you wish to use these, it is required to have git-lfs installed.

Running the pipeline

To run the pipeline, it is required to provide at least the --genotypes and --phenotypes parameters.

--genotypes

hdf5 file containing the SNP information across all accessions as a binary matrix compared to the Col-0 reference allele. This has to be stored in a certain way inside a HDF5 file and is the main reason why this pipeline only works for Arabidopsis currently. In the future, I plan to add support for more standardized formats such as PLINK or VCF, so it can be used for other organisms as well without having to create a hdf5 SNP-matrix first.

--phenotypes

Plain CSV table containing atleast 2 columns. The first column has to contain the accession ID, which will be matched with the genotype information in the SNP-matrix. The remaining trait column(s) should contain the phenotype that was measured for the respective accession ID. (See assets/phenotypes for examples)

Note: if there are more than 2 traits in the phenotype table, the pipeline will automatically split the table and submit parallel jobs for each phenotype.

If these 2 parameters are provided, the pipeline is ready to go (using default settings, test data within the repository and assuming that docker is running on the system):

nextflow gwas-nf/main.nf --phenotype gwas-nf/assets/phenotypes/FT10_mean.csv --genotype gwas-nf/assets/genotypes/regmap.hdf5 -profile docker

Other Parameters

--multitrait (default: false)

If specified, the pipeline runs in multitrait mode. In this case, 2 phenotype tables have to be provided and the path has to be enclosed in quotes. The pipeline will match them by the names of the trait columns. Example:

nextflow gwas-nf/main.nf --phenotype "gwas-nf/assets/phenotypes/{FT10,FT16}_mean.csv" --genotype gwas-nf/assets/genotypes/regmap.hdf5 --multitrait -profile docker

--maf (default: 0.05)

Minor allele frequency threshold for SNPs to be considered. Defaults to 5%.

--transform (default: 'mean_standardize')

Transformation to be applied to the phenotype vector, valid options include 'no_transformation', 'mean_standardize', 'quantile_gaussianize' and 'boxcox' Defaults to zscore scaling of the data to zero mean and unit variance.

--trait (default: false)

Option to select a single column from a large phenotype table. Has to be a string matching the column name enclosed in quotes.

--locus (default: false)

Option to specify a SNP location as a fixed effect. Needs to be passed as a comma-separated list of chromosome and position enclosed in quotes. Example:

nextflow gwas-nf/main.nf --phenotype "gwas-nf/assets/phenotypes/FT10_mean.csv" --genotype gwas-nf/assets/genotypes/regmap.hdf5 --locus '3,2761231' -profile docker

--outdir (default: ./results)

Directory where the results should be stored.

Cluster specific profiles

BioHPC Genomics cluster

The pipeline has a preconfigured profile to run on the BioHPC Genomics cluster using Charliecloud (requires Nextflow 21.03.0-edge or later)

module use /dss/dsslegfs02/pn73se/pn73se-dss-0000/spack/modules/x86_avx2/linux*
module load nextflow/21.03.0-edge charliecloud/0.22

nextflow gwas-nf/main.nf --phenotype "gwas-nf/assets/phenotypes/FT10_mean.csv" --genotype gwas-nf/assets/genotypes/regmap.hdf5 -profile biohpc_gen

CBE cluster

The pipeline has a preconfigured profile to run on CBE using Singularity

module load nextflow/20.10.0

nextflow gwas-nf/main.nf --phenotype "gwas-nf/assets/phenotypes/FT10_mean.csv" --genotype gwas-nf/assets/genotypes/regmap.hdf5 -profile cbe