snakemake pipeline for the paper 'Leveraging pleiotropy identifies common variant associations with selective IgA deficiency'

This repository contains a snakemake pipeline to generate the results presented in the paper 'Leveraging pleiotropy identifies common variant associations with selective IgA deficiency'. It has been tested with v8.23 of snakemake.

I've used conda and docker to try to make things as reproducible as possible. snakemake should pull the necessary dependencies and images managed by these two tools, but note that one can't manage installation of snakemake itself within the pipeline: I recommend conda to manage snakemake alongside the pipeline.

Note that the master branch contains the docker-enabled version; the conda branch contains a version in which the workflow is managed predominantly by conda. Each version uses conda and docker to some extent, e.g. the locuszoomr package is handled by docker in both using an image I maintain (owing to difficulties with getting conda to manage the R package's dependencies), whilst ldsc and its Python 2 environment is handled by conda (I did not want to add a conda environment to the docker image).

I think docker is preferable for managing workflows like this, but I maintain the conda version (and in fact used an earlier version of the same to produce the results in the paper) as my current HPC environment does not play nicely with the way snakemake uses apptainer (see this GitHub issue if you'd like to know more).

If you are using the conda version, I recommend running snakemake -n all_igad_paper_artifacts first to preview the workflow, then snakemake --conda-create-envs-only all_igad_paper_artifacts so that the prerequisite conda environments are built first. This can be a quite a lengthy process and snakemake doesn't log much output whilst it is happening. Note also that ldak is not provided by conda, you will have to install it and place it on the path.

Reproduction

The smk files in the directory workflow/rules/pub/igad_paper contain the snakemake rules which generate outputs presented in the paper, such as figures (figures.smk), tables (tables.smk), and data sets. Targeting snakemake at the rule all_igad_paper_artifacts in workflow/rules/pub/igad_paper/export_data.smk should have it generate all the results in the paper, although I warn you to use snakemake -n or snakemake --dry-run to get a preview of just how much processing this will take; it won't be cheap in terms of CPU hours!

Some potential issues with this pipeline:

• I used plink v2.00a6LM in the paper and whilst this version is used in the docker image, the version made available through Bioconda (and used in the conda branch) is (as of 22/10/24) v2.00a5.12
  • The docker image contains LDAK v6 wheres LDAK v5 was used to produce the results in the paper

To my knowledge, neither of these versioning discrepancies should affect the results of this workflow in any significant way.

As this pipeline has been spun out of a larger one containing work relating to several projects, it is possible that I have introduced regressions when downsizing repo content. Please open an issue if you encounter an error and I'll do my best to help you.