A powerful and efficient set test for genetic markers that handles confounders

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Set-based association tests are crucial for complex traits, aggregating weak signals and reducing multiple testing burdens.
• Existing methods often fail to account for confounding factors like family relatedness and population structure, limiting their application in large-scale genetic studies.

Purpose of the Study:

• To develop and validate a new set testing approach that effectively addresses confounding by family relatedness and population structure.
• To enhance the power and accuracy of genetic association studies for complex traits using large datasets.

Main Methods:

• A novel linear mixed model incorporating two random effects: one for set association and one for confounders.
• A computational speedup technique for two-random-effects models, enabling analysis of extremely large cohorts.
• Comparison of the likelihood ratio test (LRT) and score test within the proposed framework.

Main Results:

• The new approach successfully corrects for population structure and family relatedness in genetic association analyses.
• The likelihood ratio test demonstrated superior power while maintaining type I error control compared to the score test.
• The method identified genes not detectable by univariate analysis in a large Crohn's disease cohort.

Conclusions:

• The developed set testing method provides a robust solution for analyzing genetic associations with complex traits in the presence of confounding factors.
• This approach significantly improves the ability to detect genetic signals in large, structured datasets, advancing the field of genetic epidemiology.
• A Python library is available for implementing this advanced statistical genetics method.