Poor statistical power in population-based association study of gene interaction

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Statistical epistasis, or gene-gene interaction, describes nonadditive effects between polymorphic sites on different genes influencing a shared phenotype.
• Despite its importance in complex trait analysis, identifying statistical epistasis in genetic association studies has proven challenging, yielding limited conclusive evidence.

Purpose of the Study:

• To develop a statistical model representing statistical epistasis as an additional linkage disequilibrium between polymorphic sites of distinct risk genes.
• To evaluate and compare the statistical power for detecting gene-gene interactions under various hypothesis scenarios.

Main Methods:

• A novel statistical model was formulated to conceptualize statistical epistasis as an extended linkage disequilibrium between polymorphic sites of different risk genes.
• Statistical power calculations were performed to assess the efficacy of detecting gene-gene interactions, with comparisons made across different hypothetical situations.

Main Results:

• Statistical power for detecting gene-gene interactions positively correlates with interaction coefficient, relative risk, and linkage disequilibrium with genetic markers.
• However, the power for discovering interactions remains substantially lower compared to standard single-site association tests.
• Employing stringent statistical criteria (e.g., p-value ≤ 5.0 × 10⁻⁸) significantly hinders the identification of gene-gene interactions across most scenarios.

Conclusions:

• The observed scarcity of detected epistasis is an inherent limitation stemming from the fundamental statistical principles underlying current genetic association study methodologies.
• This difficulty in identifying gene-gene interactions appears to be an intrinsic characteristic of the research field itself, rather than a reflection of biological absence.