🐄 bovreg-twas-suite

bovreg-twas-suite is a modular Nextflow pipeline suite for performing transcriptome-wide association studies (TWAS) in cattle. It includes workflows to generate imputed genotypes and expression inputs from RNA-seq data for model training, and to run TWAS using trained models to map trait-associated genes.

It comprises two independent workflows:

Workflow	Description
bovreg-twas-preprocess	Preprocess RNA-seq and genotype data — includes trimming, alignment, quantification, imputation, and PEER factor inference
bovreg-twas-model	Train elastic net gene expression prediction models and run S-PrediXcan

---

📦 Installation

Clone the repository:

git clone https://github.com/evotools/bovreg-twas-suite.git
cd bovreg-twas-suite

Ensure you have:

• Nextflow v22.10.0 or higher
• Either: Conda, Docker, or Singularity
• Reference genome files, annotation, and optionally a GWAS SNP list

---

🛠 Workflow Overview

1️⃣ Preprocessing Workflow: bovreg-twas-preprocess

Runs:

• trim_galore, fastqc, star, featureCounts, kallisto
• automatic MultiQC aggregation of FastQC and STAR logs
• GLIMPSE phasing and imputation
• PEER factor estimation

Run example:

nextflow run workflows/bovreg-twas-preprocess \
  --fasta cattle.fa \
  --gtf annotation.gtf \
  --input_sheet samples.tsv \
  --panel_vcf reference_panel.vcf.gz \
  --map_file genetic_map.txt \
  -profile conda

Use -stub-run to validate structure.

Expected QC output:

• results/multiqc/multiqc_report.html

---

2️⃣ Model Workflow: bovreg-twas-model

Runs:

• VCF and SNP annotation filtering
• Training of elastic net models using nested CV
• Covariance matrix computation
• Merging results into SQLite
• Running S-PrediXcan
• Genome-wide TWAS Manhattan plots (one per trait result file)

Run example:

nextflow run workflows/bovreg-twas-model \
  --gwas_snps gwas_snps.txt \
  --gwas_vcf gwas.vcf.gz \
  --gtf annotation.gtf \
  --genotype_file genotypes.txt \
  --expression_file expression.txt \
  --covariates_file covariates.txt \
  --snp_annot_file snp_annot.txt \
  --gwas_sumstats_glob "sumstats/*.tsv" \
  -profile conda

Expected model plotting outputs:

• results_model/s_predixcan/s_predixcan_results_<trait>.tsv
• results_model/twas_plots/twas_manhattan_<trait>.png
• results_model/twas_plots/twas_plot_manifest_<trait>.tsv

---

🔎 Parameters

Each workflow includes a nextflow_schema.json so you can explore parameters with:

nextflow run workflows/bovreg-twas-preprocess --help
nextflow run workflows/bovreg-twas-model --help

---

🧪 Testing

You can validate the full structure using:

nextflow run workflows/bovreg-twas-preprocess -stub-run
nextflow run workflows/bovreg-twas-model -stub-run

Linux cluster testing (SGE + conda)

Test templates are provided in tests/preprocess/. Update tests/preprocess/samples.test.tsv and tests/preprocess/params.test.json with valid cluster paths before running.

Validate environment and profile resolution:

nextflow -version
nextflow config workflows/bovreg-twas-preprocess -profile sge

Structural (no compute-heavy execution):

nextflow run workflows/bovreg-twas-preprocess \
  -profile sge \
  -stub-run \
  -params-file tests/preprocess/params.test.json

Short real test run:

nextflow run workflows/bovreg-twas-preprocess \
  -profile sge \
  -params-file tests/preprocess/params.test.json \
  --queue <queue_name> \
  --max_cpus 8 \
  --max_memory '32 GB' \
  --max_time '8.h' \
  -resume

MultiQC verification checklist:

• Confirm SGE jobs are submitted and completed.
• Confirm fastqc and star_align tasks both complete.
• Confirm multiqc runs and publishes:
  • results/multiqc/multiqc_report.html
• Open the report and verify FastQC and STAR sections are present.
• Re-run with -resume and verify completed tasks are reused.

---

📁 Folder Structure

workflows/
├── bovreg-twas-preprocess/
│   ├── main.nf
│   ├── modules/
│   ├── nextflow.config
│   └── nextflow_schema.json
├── bovreg-twas-model/
│   ├── main.nf
│   ├── modules/
│   ├── scripts/
│   ├── nextflow.config
│   └── nextflow_schema.json
configs/           → Profiles for HPC or containerised execution
envs/              → Conda YAMLs for each module group
bin/               → Custom scripts (e.g., parse_sample_sheet.py)

---

✨ Authors

Developed by Siddharth Jayaraman for the BovReg project.
Includes contributions from Roslin Institute pipelines and PredictDB modeling strategies.

---

📝 License

MIT License. See LICENSE for full text.