PennPRS

PennPRS enables efficient polygenic risk score (PRS) training using pseudo-training methods on a scalable cloud computing platform, allowing users to train PRS models without requiring individual-level genetic data.

Key Features:

• Pseudo-training algorithms: Novel pseudo-training algorithms enable model training for multiple PRS methods and ensemble approaches without requiring individual-level data.
• Single- and multi-ancestry PRS training: The platform supports online single-ancestry and multi-ancestry PRS training with seven methods.
• GWAS Catalog integration: Users can upload their own data or query from more than 27,000 datasets in the GWAS Catalog.
• Pre-trained PRS models: PRS models for over 8,000 phenotypes have been trained using the pseudo-training pipeline and their PRS weights are publicly accessible.

Scientific Applications:

• Precision medicine risk prediction: PRS model training for risk prediction and stratification in precision medicine.
• Large-scale phenotype analysis: Validation and application across over 6,000 phenotypes through extensive simulations and large-scale real data analyses across various data sources.

Methodology:

PennPRS employs novel pseudo-training algorithms for multiple PRS methods and ensemble approaches, validated through extensive simulations and large-scale real data analyses.