NeuPred

NeuPred implements a Bayesian regression framework to construct polygenic risk scores (PRS) from GWAS summary statistics and to optimize chromosome-level priors via summary-statistics-based cross-validation for improved prediction of complex disease genetic risk.

Key Features:

• Bayesian Regression Framework: Employs a data-adaptive Bayesian regression model with flexible prior choices to accommodate varying genetic architectures across chromosomes.
• Chromosome-Level Prior Selection: Uses a summary-statistics-based cross-validation strategy to automatically select chromosome-level priors and reveal variability in prior preferences across chromosomes.
• Improved Predictive Accuracy: Demonstrates substantial improvements in predictive r² in simulations and real-world applications using the Wellcome Trust Case Control Consortium cohort and large-scale genome-wide association studies compared to existing methods.
• Computational Efficiency: Maintains computational efficiency comparable to or better than current state-of-the-art Bayesian methods.

Scientific Applications:

• Genetic epidemiology: Enables construction and evaluation of PRS for studying genetic contributions to complex diseases using GWAS summary statistics.
• Personalized medicine and public health: Supports risk prediction for a wide range of complex diseases to inform personalized medicine strategies and population-level health analyses.

Methodology:

Constructs PRS using a Bayesian regression framework that integrates GWAS summary statistics and applies summary-statistics-based cross-validation to select chromosome-level priors.