Polygenic Risk Score (PRS) Pipeline

This pipeline is a reproducible workflow for calculating and validating PRS for Type 2 Diabetes (T2D) across the UK Biobank, All of Us, and Lyday datasets. This pipeline leverages Python, R, and Nextflow to enable robust PRS analysis with ancestry adjustment and statistical validation.

Repo Structure

Folder	Description
Data/	Describes each cohort: Lyday, All of Us, and UK Biobank. Includes info on preprocessing.
Tools/	Contains usage notes for core tools like pgsc_calc, PRSice-2, PRScs, Beagle, and more.
Ancestry Adjustment/	Focuses on ancestry-normalized PRS and includes the logic replicated from the pgsc_calc framework.
Validation/	Scripts and results for statistical validation using R.
PRS_Pipeline/	(Coming soon) Python-based CLI pipeline to modularly run each step with checkpointing.

Datasets

• UK Biobank: Used for testing and benchmarking PRS tools (data already imputed and standardized)
• All of Us: PRS calculation in addition to ancestry-adjustment using pgsc_calc
• Lyday: Raw VCFs requiring genotype harmonization, imputation, and conversion prior to PRS calculation

Each dataset has its own documentation inside the Data/ directory.

Tools Utilised

• pgsc_calc: Nextflow-based official PRS scoring pipeline
• PRSice-2: Clumping + thresholding tool
• PRScs: Bayesian regression method for PRS
• Beagle 5.5: Imputation
• conform-gt: Genotype harmonization
• Python for pipeline scripting
• R for statistical validation (e.g., AUC, Odds Ratio, Cohen’s d)

Getting Started

This project is under development. To reproduce or test:

1. Navigate to the dataset of interest under Data/
2. Set up the tools required under Tools/
3. Run the PRS pipeline using pgsc_calc or the custom Python workflow
4. Validate the scores using the scripts in Validation/

References

• Lambert et al. (2021), The Polygenic Score Catalog as an open database for reproducibility and systematic evaluation.
• All of Us Research Program, NIH

Maintained by

Valli Sree Lasya Pasumarthy ([email protected])
Graduate Research – Jordan Lab
Masters of Science in Bioinformatics
School of Biological Sciences, Georgia Institute of Technology