An efficient resampling method for calibrating single and gene-based rare variant association analysis in...

Area of Science:

• Genetics
• Statistical Genetics
• Computational Biology

Background:

• Genetic association studies often analyze variants with low minor allele counts (MACs), especially when focusing on deleterious variants.
• Standard asymptotic tests struggle with calibration and power for binary phenotypes when MAC is low, leading to unreliable results.
• Conventional resampling methods for empirical p-values are computationally intensive and can be conservative.

Purpose of the Study:

• To develop a computationally efficient resampling method for variant aggregation tests that addresses calibration issues in low MAC scenarios.
• To improve the speed and reliability of score-based association tests, particularly for rare variants.
• To provide a calibrated framework for evaluating genetic association results using empirical p-values.

Main Methods:

• Developed an efficient resampling strategy for single and multiple variant score-based tests, leveraging the contribution of individuals with minor alleles.
• Incorporated mid-p-values to mitigate the conservativeness of resampling results.
• Utilized estimated minimum achievable p-values for QQ plot calibration and determining the effective number of tests.

Main Results:

• Achieved over 1000-fold improvement in computational efficiency compared to conventional resampling for low MAC variant sets.
• Demonstrated well-calibrated results in a deep exome sequence case-control study.
• Significantly reduced computation time while maintaining statistical rigor.

Conclusions:

• The proposed efficient resampling method offers a computationally feasible and statistically robust alternative for genetic association testing with low MAC variant sets.
• This approach enhances the reliability and efficiency of analyzing rare variants in large-scale genetic studies.
• The method provides a calibrated framework for interpreting association results and controlling the false discovery rate.