Overcoming the "feast or famine" effect: improved interaction testing in genome-wide association studies

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genetic association analysis of complex traits aims to identify gene-gene (GxG) and gene-environment (GxE) interactions.
• Genome-wide interaction studies (GWIS) face challenges like multiple comparisons and heteroscedasticity.
• A significant issue in GWIS is the "feast or famine" effect, where Type 1 error rates fluctuate unpredictably.

Purpose of the Study:

• To identify and explain the "feast or famine" effect in GxE GWIS.
• To develop methods for detecting and correcting this effect.
• To improve the reliability and power of interaction detection in GWIS.

Main Methods:

• Theoretical analysis to pinpoint the cause of the "feast or famine" effect (conditioning on variables).
• Development of a diagnostic ratio to identify affected GWASs.
• Introduction of the TINGA method to adjust interaction test statistics for uniform p-values under the null hypothesis.
• Utilized linear mixed models (LMM) to account for covariates, population structure, and heteroscedasticity.

Main Results:

• The "feast or famine" effect leads to under- or overinflated p-values, causing false positives, reduced power, and inconsistent findings.
• The diagnostic ratio effectively detects GWASs susceptible to this effect.
• TINGA demonstrated Type 1 error control and improved power in simulations.
• TINGA was applied to study epistasis in *Arabidopsis thaliana* flowering time.

Conclusions:

• The "feast or famine" effect is a critical, previously underappreciated challenge in GWIS.
• TINGA provides a robust solution for accurate and powerful interaction detection in GxE studies.
• This method enhances the reproducibility and validity of findings from genome-wide interaction analyses.