A unified model based multifactor dimensionality reduction framework for detecting gene-gene interactions.

Area of Science:

• Genetics and Bioinformatics
• Statistical Genomics

Background:

• Gene-gene interactions (GGI) are crucial for understanding complex traits but are challenging to detect.
• Existing multifactor dimensionality reduction (MDR) methods struggle with significance evaluation and computational burden.
• Current MDR approaches may miss causal epistasis due to masking by marginal effects.

Purpose of the Study:

• To develop an efficient and robust method for detecting GGI in genome-wide association studies.
• To address the limitations of conventional MDR methods, including computational intensity and difficulty in assessing multi-locus model significance.
• To propose a unified model-based MDR approach (UM-MDR) that can handle various trait types and evaluate existing MDR extensions.

Main Methods:

• A two-step unified model-based MDR approach (UM-MDR) is proposed.
• Significance of multi-locus models is assessed using a regression framework with a semi-parametric correction procedure.
• This approach avoids computationally intensive permutations for significance testing and controls Type I error rates.

Main Results:

• UM-MDR demonstrates comparable or superior power to traditional MDR in simulations.
• The proposed method effectively detects causal epistasis, especially when marginal effects are present.
• Simulation studies and real data analysis confirm the utility and efficiency of UM-MDR.

Conclusions:

• UM-MDR offers an efficient and powerful alternative to existing MDR methods for GGI analysis.
• Its flexibility in handling different trait types and its computational efficiency make it a valuable supplement.
• The method successfully addresses limitations in significance testing and detection of epistasis in complex trait studies.