Rat TWAS Hub

Code to generate the data and website for Rat TWAS Hub.

This repository runs large-scale, multi-modal TWAS analyses in rat and produces static Markdown reports plus a Jekyll site. It is adapted from the (human) TWAS HUB and extends gene expression-based TWAS to multi-modal RNA phenotypes (xTWAS) using Pantry to generate phenotypes for the transcriptome models.

Requirements

• Python 3 with snakemake and pandas.
• R (4.x recommended) with packages:
  • required: tidyverse, yaml, biomaRt, optparse, plink2R, RColorBrewer, coloc
• GNU core tools (awk, sed, join, tar, gzip) and parallel.
• Ruby + bundler for Jekyll site builds (see jekyll/Gemfile).
• rsync + ssh for deployment (optional).

Notes:

• The cross-species step uses Ensembl via biomaRt and requires network access.
• TWAS runs require binary PLINK reference panels (see inputs below).

Inputs and parameter files

The pipeline expects the following inputs under data/:

• panels.tsv: Transcriptome panel metadata (columns include PANEL, TISSUE, MODALITY, N).
• traits.tsv: Trait metadata (columns include OUTPUT, ID, N, PROJECT, TAGS, NAME, DESCRIPTION).
• projects.tsv: Project metadata used for the site.
• sequence_report.tsv: Chromosome sizes for porcupine plots.
• WEIGHTS/<tissue>/<modality>.pos and <modality>.profile: FUSION profiler outputs used to build data/all_models.tsv.
• LDREF/Brain_v4.chr{chrom}.{bed,bim,fam}: PLINK reference LD panels used by FUSION.
• gwas_original/sumstats/{project}/regressedlr_{trait}_chrgwas{chrom}.mlma.gz: GWAS summary statistics (GCTA output).
• gwas_original/traits.tsv: Project + trait metadata used by scripts/prepare/prepare_trait_table.R (columns: project_id, trait_id, tags, name, description).
• cross_species/genes.models.nfo and cross_species/genes.nfo: Human reference lists for the cross-species table.

Generated outputs live in:

• data/all_models.tsv, data/twas_out/, data/traits.summary.tsv, data/genes_n_models.tsv, data/genes_n_assoc.tsv
• jekyll/ (traits, genes, and _data assets consumed by the site)

Workflow

1. Prepare trait metadata (optional, if starting from raw GWAS):

scripts/prepare/prepare_traits.sh

This reads data/gwas_original/traits.tsv and GCTA logs to produce data/traits.tsv.

2. Run TWAS + post-processing + site data build:

snakemake -j 8

If running on HPC, be sure to run in a job, or have snakemake run steps as jobs using, e.g., snakemake-executor-plugin-slurm.

3. Build the Jekyll site:

cd jekyll
bundle install
bundle exec jekyll build

4. Deploy (optional, requires SSH access to the target host):

scripts/site/deploy.sh

Repository layout

• Snakefile: Main workflow (TWAS, post-processing, and site data generation).
• scripts/twas/: FUSION TWAS and post-processing scripts.
• scripts/site/: Jekyll data builders and plot generation.
• scripts/prepare/: Trait table helpers for GWAS inputs.
• jekyll/: Site templates and assets.