BayesKAT

BayesKAT detects genetic associations between traits and groups of preassigned genetic features by applying a Bayesian kernel-based association test that adaptively selects optimal kernel functions to capture polygenic and non-linear dependencies relevant to complex diseases.

Key Features:

• Kernel-based association test: Uses kernel methods to test group-level associations between traits and sets of genetic features.
• Bayesian kernel selection: Employs Bayesian methodology to automatically select the optimal kernel function from the data.
• Adaptive composite kernel: Selects a composite kernel adaptively to improve robustness across different genetic architectures.
• Type I error control and power: Designed to maintain stringent type I error control while improving statistical power for detecting group-level associations.
• High-dimensional data handling: Operates effectively on high-dimensional genotype datasets common in modern genetics research.
• Integration of prior biological knowledge: Incorporates prior knowledge derived from experiments or known mechanisms to inform association tests.
• Validation on simulated and real data: Performance has been evaluated through comparisons using simulated and large-scale real-world genetics datasets.
• Targets group-level features: Applied to predefined groups such as biological pathways, co-expression gene modules, and protein complexes.
• Addresses GWAS limitations: Specifically targets polygenic influences and non-linear dependencies that standard GWAS single-SNP analyses may miss.

Scientific Applications:

• Group-level association discovery: Detects associations between traits and groups of preassigned genetic features such as pathways and complexes.
• Polygenic architecture analysis: Identifies polygenic and non-linear genetic influences underlying complex traits and diseases.
• Functional module interrogation: Tests sets like co-expression gene modules and protein complexes for trait associations.
• Mechanistic insight generation: Integrates prior biological information to support mechanistic interpretation of genetic associations.
• Method comparison and validation: Evaluates performance against other methods using simulated and large-scale real genetics data.

Methodology:

Implements a kernel-based association test with Bayesian automatic selection of optimal and composite kernels from the data, incorporates prior biological knowledge, handles high-dimensional genotype data, and evaluates performance via comparisons on simulated and real-world large-scale genetics datasets.