Users Manual

Installation

  1. Before installation, we highly recommend creating a new environment using conda so that it will not affect the software versions of other projects. For example, run:

    conda create -n env-mrpeg python=3.10

  2. If you are using a Mac with an Apple M1 or newer chip, you should initialize conda using miniforge to ensure compatibility (see this link for a previous issue). On most HPC systems, this is usually not necessary.

  3. Finally, download the latest repository and install via pip:

    git clone https://github.com/gusevlab/mrpeg.git
cd mrpeg
pip install .

Data Preparation

mrpeg requires users to input GWAS summary data, eQTL summary data, and gene-to-gene effects matrix to perform inference.

Users can check testing data in Testing Data section to see the format requirement.

Testing Data

We provide example data in ./data/ folder to test out mrpeg.

We simulated GWAS summary and eQTL summary data using real cis-eQTLs identified in eQTLGen whose genotypes are from 1000G project.

Using ./data/make_example.py, we simulated GWAS, eQTL summary data. Users can run the script in ./data/run_example.sh.

Usage

mrpeg software is very easy to use.

Performing inference

cd ./data/
mrpeg peg --gwas example_gwas.tsv.gz \
  --eqtl example_eqtl.tsv.gz \
  --perturb example_perturbation.tsv.gz \
  --gwas-cols chrom snp a1 a0 beta se \
  --eqtl-cols chrom snp a1 a0 z gene \
  --ref-geno plink/geno_chr\* \
  --trait "mediating_gene" \
  --top-signal 1 \
  -o tmp_results_mediating_gene

Gene annotation

We also implement a function for gene annotation to compute the GWAS signals given gene annotations:

cd ./data/
mrpeg signal --gwas example_gwas.tsv.gz \
  --gwas-cols chrom snp pos beta se \
  --ref ref_gene_info.tsv.gz \
  --ref-cols CHR P_MID_FLANK0 P_MID_FLANK1 ID2 \
  --window 0 \
  --trait example \
  --chr 1 \
  -o tmp_results_signal