Identifying gene-gene interactions using penalized tensor regression.

Area of Science:

• Genetics
• Computational Biology
• Biostatistics

Background:

• Gene-gene (G×G) interactions are crucial for complex disease development, but traditional marginal analyses are limited.
• Existing joint analysis methods face computational challenges due to the "main effects, interactions" hierarchical constraint.

Purpose of the Study:

• To propose a new, computationally efficient method for identifying significant G×G interactions using joint modeling.
• To address the limitations of current methods in handling high-dimensional genetic data and complex disease mechanisms.

Main Methods:

• Utilizing tensor regression to manage high-dimensional data.
• Employing a penalization technique for feature selection.
• Leveraging the natural hierarchical structure accommodation of tensor regression for simpler optimization.

Main Results:

• The proposed method demonstrates superior performance compared to alternatives in simulation studies.
• Analysis of The Cancer Genome Atlas (TCGA) lung cancer and melanoma data identified significant markers.
• The method achieved better prediction performance on real-world cancer datasets.

Conclusions:

• The novel tensor regression approach offers an efficient and effective solution for identifying G×G interactions in complex diseases.
• This method enhances understanding of genetic contributions to diseases like cancer and improves predictive modeling.
• It provides a computationally feasible alternative for large-scale genetic interaction studies.