RapidoPGS

RapidoPGS computes polygenic scores from summary-level genome-wide association study (GWAS) data to estimate individual genetic predisposition to complex traits.

Key Features:

• No Need for Validation Dataset: Internally tunes parameters using only summary-level GWAS data, removing the requirement for an independent validation dataset.
• Computational Efficiency: Provides substantial speed improvements relative to LDpred2, PRScs, and SBayesR while maintaining competitive predictive performance.
• No Requirement for LD Matrices: Bypasses computation of linkage disequilibrium (LD) matrices by leveraging fine-mapping principles.
• Adaptive Shrinkage of Effect Sizes: Calculates the posterior probability that each genetic variant is causal and uses these probabilities to adaptively shrink effect sizes based on LD and association strength.
• Performance: In case-control datasets shows median r² differences of −0.0092 and −0.0042 relative to LDpred2 and PRScs, respectively, while achieving up to 17,000-fold speed increases.
• Implementation: Implemented in R and accepts user-supplied summary statistics or automatic download from the GWAS catalog.

Scientific Applications:

• Large-scale genetic studies: Supports rapid computation of PGS for large-scale GWAS and meta-analyses.
• Polygenic risk prediction: Facilitates estimation of individual genetic predisposition to complex traits, including case-control phenotypes.
• Epidemiology: Enables fast genetic risk scoring across populations for epidemiological analyses.
• Evolutionary biology and genetic-architecture studies: Incorporates fine-mapping-informed causal probabilities to inform analyses of genetic architecture and evolutionary questions.

Methodology:

Operates on summary-level GWAS data with internal parameter tuning (no validation dataset), leverages fine-mapping principles to compute posterior probabilities of causality for each variant, uses those probabilities to adaptively shrink effect sizes based on LD and association strength, bypasses explicit LD matrix computation, and is implemented in R with options to use user-supplied summary statistics or download from the GWAS catalog.