Flexible Bayesian Inference for Identifying Significantly Correlated Multiple Pathway Sets

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Pathway-based analysis in genetics is crucial for detecting subtle expression changes.
• Interactions among biological pathways complicate marginal analysis, leading to potential errors.
• Existing methods often fail to account for inter-pathway dependencies in clinical outcome studies.

Purpose of the Study:

• To develop a flexible Bayesian inference method for identifying significantly correlated high-dimensional functions (pathways) with a response variable.
• To address the challenge of unknown and complex relationships due to inter-pathway dependence.
• To improve the accuracy of genetic pathway analysis by accounting for pathway interactions.

Main Methods:

• Proposed a generalized fused kernel machine regression approach.
• Developed a data-driven, flexible Bayesian inference framework.
• Utilized Bayes factor for multiple testing adjustment, accommodating dependence through a flexible structure.

Main Results:

• The proposed method effectively identifies significantly correlated high-dimensional functions with continuous or binary response variables.
• Bayesian inference with Bayes factor adjustment successfully accommodates inter-pathway dependence.
• Demonstrated benefits through simulation studies and analysis of genetic pathway data for type II diabetes.

Conclusions:

• The flexible Bayesian inference method offers a robust approach for analyzing high-dimensional functions in complex biological systems.
• Accounting for pathway interactions is essential for accurate identification of significant functions and reliable clinical outcome prediction.
• The method provides a valuable tool for genetic pathway analysis, particularly in complex diseases like type II diabetes.