A powerful association test of multiple genetic variants using a random-effects model

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Emerging interest in sequencing-based association studies for multiple rare variants.
• Limitations of existing methods like collapsing and SKAT (Sequence Kernel Association Test) due to missing genotypes and variant types.

Purpose of the Study:

• Introduce a novel association test, T REM, designed to overcome limitations of current methods.
• Evaluate the performance of T REM against competing tests under various genetic and data conditions.

Main Methods:

• Developed T REM based on a random-effects model, allowing for missing genotypes and simultaneous analysis of common and rare variants without requiring specific weighting functions.
• Conducted extensive simulations to assess type I error rates and statistical power under diverse scenarios (sample size, missingness, variant frequency, effect directionality).

Main Results:

• T REM demonstrated valid type I error rates and was less sensitive to non-causal variants and missing genotypes compared to other tests.
• The T REM test exhibited superior power performance when variant effects were consistent in direction.
• Application to the Shanghai Breast Cancer Study identified rare causal variants at the FGFR2 gene, with T REM yielding more consistent results across variant sets.

Conclusions:

• T REM is a robust and powerful statistical test for sequencing-based association studies, effectively handling missing data and mixed variant frequencies.
• The method offers improved consistency and power, making it a valuable tool for genetic association research, particularly in complex diseases.