An efficient Bayesian meta-analysis approach for studying cross-phenotype genetic associations

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Pleiotropy, where a single genetic locus influences multiple traits, is crucial for understanding shared genetic susceptibility.
• Identifying specific traits associated with a pleiotropic locus is challenging but essential for biological interpretation.
• Existing methods may not fully capture the complexity of aggregate-level pleiotropic associations.

Purpose of the Study:

• To introduce CPBayes, a novel Bayesian meta-analysis approach for simultaneous analysis of genetic associations across multiple phenotypes.
• To accurately measure aggregate-level pleiotropic association evidence and identify optimal subsets of traits linked to a risk locus.
• To provide a flexible method applicable to diverse study designs and trait data.

Main Methods:

• Utilizes a unified Bayesian statistical framework with a spike and slab prior.
• Employs Markov Chain Monte Carlo (MCMC) via Gibbs sampling for a fully Bayesian analysis.
• Accounts for heterogeneity in genetic effect size and direction across traits.

Main Results:

• Simulations demonstrate CPBayes' superior accuracy in selecting traits underlying pleiotropic signals compared to ASSET.
• Genome-wide analysis of 22 traits in the Kaiser GERA cohort identified six independent pleiotropic loci.
• A significant locus at 1q24.2 was associated with five diseases: Dermatophytosis, Hemorrhoids, Iron Deficiency, Osteoporosis, and Peripheral Vascular Disease.

Conclusions:

• CPBayes offers a robust and accurate method for dissecting pleiotropic associations from summary-level data.
• The method enhances the identification of specific trait contributions to complex genetic loci.
• An R-package 'CPBayes' is available, facilitating the application of this advanced statistical approach in genetic research.