An optimal kernel-based U-statistic method for quantitative gene-set association analysis.

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Single-variant genome-wide association studies (GWAS) identify genetic variants for complex traits but have limitations.
• These limitations include weak marginal signals and the neglect of complex genetic variant interactions.
• Set-based association studies evaluate multiple variants jointly (e.g., within a gene or pathway) to capture systematic effects.

Purpose of the Study:

• To develop an efficient and statistically robust method for set-based association studies.
• To address the limitations of existing kernel-based testing (KBT) methods, such as inflated Type 1 error rates and computational inefficiency.
• To improve the power of detecting genetic associations for complex traits by considering variant interactions.

Main Methods:

• Proposed a novel maximum kernel-based U-statistic method within the kernel-based testing (KBT) framework.
• Utilized asymptotic results for high-dimensional settings, enabling analysis where variant number exceeds sample size.
• Developed a procedure that efficiently controls Type 1 error rates.

Main Results:

• The proposed method demonstrated effective control of Type 1 error rates.
• Achieved statistical power comparable to the optimal kernel within the candidate set.
• Showcased efficiency in handling high-dimensional genetic data where the number of variants is substantially larger than the sample size.

Conclusions:

• The developed maximum kernel-based U-statistic method offers an efficient and powerful approach for set-based association studies.
• This method provides a significant advancement over existing techniques, particularly for complex traits and large genetic datasets.
• Simulation and real data analyses confirm the method's advantages in controlling errors and enhancing detection power.