CPBayes

CPBayes performs Bayesian meta-analysis of summary-level genetic association data to detect pleiotropic loci and identify the subset of phenotypes associated with each risk locus.

Key Features:

• Unified Bayesian framework: Employs a spike and slab prior to model association status and effect sizes, distinguishing significant genetic effects from noise.
• Markov Chain Monte Carlo (Gibbs sampling): Implements fully Bayesian inference via Gibbs sampling to explore complex posterior distributions.
• Heterogeneity modeling: Accounts for heterogeneity in both the magnitude and direction of genetic effects across different traits.
• Summary-level data support across study designs: Operates on summary statistics from cohort data and separate studies with overlapping or non-overlapping subjects.
• Joint pleiotropy testing and trait selection: Simultaneously assesses aggregate-level pleiotropic association and determines the optimal subset of associated traits at a risk locus.
• Improved trait-selection accuracy: Simulation studies report higher accuracy in selecting associated traits underlying a pleiotropic signal compared with the subset-based meta-analysis ASSET.

Scientific Applications:

• Cross-phenotype GWAS meta-analysis: Discovery of genetic loci that influence multiple phenotypes using summary statistics from GWAS.
• Pleiotropy mapping and interpretation: Identification of shared genetic susceptibilities to elucidate the genetic architecture of diseases and traits.
• Kaiser GERA cohort example: Application to 22 traits identified six independent pleiotropic loci, including a locus at 1q24.2 associated with Dermatophytosis, Hemorrhoids, Iron Deficiency, Osteoporosis, and Peripheral Vascular Disease.

Methodology:

Integrates summary-level data into a unified Bayesian model using a spike and slab prior and performs inference with Gibbs sampling, accounting for heterogeneity in effect magnitude and direction while assessing aggregate pleiotropic association and selecting associated trait subsets.