Gene-Based Testing of Interactions Using XGBoost in Genome-Wide Association Studies

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) are crucial for identifying genetic variants associated with diseases.
• Detecting gene-gene interactions is vital for a comprehensive understanding of complex traits.
• Existing gene-based interaction methods have limitations in statistical detection due to assumptions on single nucleotide polymorphism-trait associations.

Purpose of the Study:

• To propose a novel gene-based method, GGInt-XGBoost, for detecting gene-gene interactions.
• To overcome the limitations of current methods in statistical detection and biological interpretability.
• To provide a statistically robust approach for identifying significant gene-gene interactions.

Main Methods:

• Developed GGInt-XGBoost, a method leveraging XGBoost for gene-based interaction detection.
• Assumed an additive relationship for log odds ratio of disease traits in the absence of gene-gene interactions.
• Utilized the difference in XGBoost model error (with and without additive constraint) to infer interactions.
• Employed a permutation-based statistical test to assess the significance of detected interactions and provide p-values.

Main Results:

• GGInt-XGBoost demonstrated superior performance in detecting gene-gene interactions compared to previous methods.
• The method showed effectiveness on both simulated and real genetic datasets.
• The permutation-based test successfully provided statistically significant p-values for interactions.

Conclusions:

• GGInt-XGBoost offers a powerful and biologically interpretable approach for identifying gene-gene interactions.
• The proposed method enhances the statistical detection capabilities in genome-wide association studies.
• GGInt-XGBoost represents a significant advancement in the field of genetic interaction analysis.