Hierarchical Bayesian model for rare variant association analysis integrating genotype uncertainty in human sequence...

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• Next-generation sequencing (NGS) enables the study of rare genetic variants (RVs) potentially explaining complex disease heritability.
• Existing rare variant association methods often overlook sequencing errors, which can reduce statistical power and accuracy.
• Sequencing errors, mimicking rare alleles, can lead to false negatives in association tests.

Purpose of the Study:

• To develop a robust statistical framework for rare variant association analysis that explicitly accounts for sequencing errors.
• To enhance the power and accuracy of detecting associations between rare variants and complex diseases.
• To provide a flexible model capable of handling various types of rare variants and measurement errors.

Main Methods:

• Developed a hierarchical Bayesian approach integrating misclassification probability with Bayesian variable selection.
• Incorporated imputation uncertainty and minor allele frequency (MAF) into the framework.
• Utilized shrinkage-based Bayesian variable selection for robust model estimation.

Main Results:

• Demonstrated that sequencing errors significantly affect association test findings.
• The proposed Bayesian model leverages sequencing error information to improve statistical power in simulated and real data.
• Outperformed existing methods like the Sequence Kernel Association Test (SKAT) in detecting associations.
• Identified a known rare variant (FH North Karelia in LDLR) in a Finnish LDL cholesterol study, missed by SKAT and Granvil.

Conclusions:

• The developed hierarchical Bayesian method effectively addresses sequencing errors in NGS data for rare variant association studies.
• This approach offers improved statistical power and accuracy compared to existing methods, particularly in the presence of sequencing errors.
• The model's ability to detect previously missed associations highlights its utility in complex disease genetics research.