Association studies with imputed variants using expectation-maximization likelihood-ratio tests.

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genotype imputation is crucial in genetic studies, but low-quality imputation of low-frequency variants poses challenges.
• Growing reference panels improve imputation for many markers, yet increase the number of poorly imputed ones.

Purpose of the Study:

• To develop novel methods for incorporating genotype imputation uncertainty into association analyses.
• To enhance statistical power and computational efficiency in genetic association studies.

Main Methods:

• Developed an expectation-maximization likelihood-ratio test (EM-LRT) for association using posterior genotype probabilities (Scenario I).
• For imputed dosages (Scenario II), sampled genotype probabilities from their posterior distribution and applied EM-LRT.
• Validated methods through simulations and applications to real genetic datasets.

Main Results:

• Proposed EM-LRT methods maintain Type I error rates under both scenarios.
• EM-LRT-Prob (Scenario I) demonstrates optimal statistical power across various minor allele frequencies (MAF) and imputation qualities.
• EM-LRT-Dose (Scenario II) matches EM-LRT-Prob's power and surpasses standard methods for low MAF/imputation quality markers.

Conclusions:

• The proposed EM-LRT methods effectively handle imputation uncertainty in genetic association studies.
• These methods offer improved power and efficiency, particularly for challenging markers.
• Validates the utility of incorporating imputation uncertainty for robust genetic discoveries.