Pathway-guided identification of gene-gene interactions

Area of Science:

• Genetics and Bioinformatics
• Computational Biology
• Statistical Genomics

Background:

• Gene-gene interactions (GxG) are crucial for understanding complex traits.
• Existing GxG methods often focus on limited gene sets, posing challenges for large-scale analyses.
• Identifying epistasis at the gene level offers amplified signals compared to marker-marker interactions.

Purpose of the Study:

• To develop a novel regression model for detecting gene-gene interactions (GxG) within large gene lists.
• To incorporate biological pathway information and trait data for improved interaction detection.
• To provide a method applicable to complex traits and situations with sample sizes smaller than the number of predictors.

Main Methods:

• A pathway-guided regularized regression model was proposed.
• Principal components were used to summarize single nucleotide polymorphism (SNP)-SNP interactions within gene pairs.
• An L1 penalty with adaptive weights, informed by biological pathways and trait data, was employed to identify significant main and interaction effects.

Main Results:

• The proposed method demonstrated improved performance compared to approaches lacking pathway and trait guidance.
• Simulations and real data analysis validated the utility and effectiveness of the new approach.
• The method successfully identified important gene-gene interactions, offering insights for biological hypothesis generation.

Conclusions:

• The developed pathway-guided regression model offers a robust framework for exploring GxG in complex traits.
• This approach effectively integrates biological knowledge and data-driven adaptiveness for enhanced genetic interaction discovery.
• The method facilitates the formulation of testable biological hypotheses for further molecular investigation.