JBASE: Joint Bayesian Analysis of Subphenotypes and Epistasis.

Area of Science:

• Genetics
• Computational Biology
• Statistical Genetics

Background:

• Genome-wide association studies (GWAS) identify genotype-phenotype associations but explain limited heritability.
• Missing heritability in complex diseases is partly due to unmodeled genetic interactions (epistasis) and phenotypic heterogeneity.

Purpose of the Study:

• To propose JBASE (joint Bayesian analysis of subphenotypes and epistasis), an integrative model to address missing heritability.
• To explore epistasis and phenotypic heterogeneity as key factors in complex disease genetics.

Main Methods:

• Developed JBASE, an integrative mixture model for joint Bayesian analysis of subphenotypes and epistasis.
• Implemented JBASE in C++ for Linux systems.

Main Results:

• JBASE accurately identifies subphenotypes, associated variants, and their interactions in simulations.
• JBASE demonstrated superior performance (higher power, lower Type 1 error) compared to state-of-the-art methods in handling phenotypic heterogeneity.
• Applied to Type 2 diabetes data, JBASE discovered two novel epistatic modules defining BMI and waist-to-hip ratio subphenotypes, replicated in an independent dataset.

Conclusions:

• JBASE effectively addresses missing heritability by integrating subphenotyping and epistasis analysis.
• The method successfully identified and replicated novel genetic findings for Type 2 diabetes subphenotypes.