Detecting Multivariate Gene Interactions in RNA-Seq Data Using Optimal Bayesian Classification.

Area of Science:

• Bioinformatics
• Computational Biology
• Statistical Genetics

Background:

• Differential gene expression (DGE) analysis is standard for RNA-Seq data, identifying genes with significant expression differences across phenotypes.
• Current methods primarily focus on individual gene expression levels, potentially missing complex gene interactions.

Purpose of the Study:

• To extend DGE testing to a classification framework for detecting multivariate gene interactions.
• To identify novel gene interaction pairs relevant to specific phenotypes using a new computational approach.

Main Methods:

• Development of a novel computational framework integrating a hierarchical statistical model of the RNA-Seq pipeline.
• Implementation of an optimal Bayesian classifier (OBC) for gene interaction analysis.
• Utilizing Markov Chain Monte Carlo (MCMC) sampling and Monte Carlo integration for complex quantity computation.

Main Results:

• Identification of gene pairs with low classification error, indicating significant interactions, in a dietary intervention study dataset.
• Discovery of gene pairs not detected by conventional DGE methods, highlighting the framework's ability to find novel interactions.
• Demonstration of the framework's performance on a real-world gene expression dataset.

Conclusions:

• The novel OBC classification framework effectively identifies complex gene interactions from RNA-Seq data.
• This approach complements traditional DGE analysis by uncovering previously undetected gene relationships.
• A publicly available open-source software package for OBC classification on RNA-Seq data has been released.