Bayesian latent variable collapsing model for detecting rare variant interaction effect in twin study.

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• Rare genetic variants (RVs) are increasingly recognized for their role in complex phenotypes.
• Existing statistical models struggle to detect RV interaction effects due to low allele frequencies.
• Advanced methods are needed to analyze RV associations and interactions in large datasets.

Purpose of the Study:

• To propose a novel statistical method, the hierarchical Bayesian latent variable collapsing method (BLVCM), for detecting rare variant interaction effects.
• To address the challenges posed by scarce minor allele occurrences in genetic association studies.
• To provide a flexible framework applicable to various study designs, including twin studies.

Main Methods:

• Developed a hierarchical Bayesian latent variable collapsing method (BLVCM).
• Parameterizes RV signals using latent variables within a Bayesian framework.
• Designed for twin data analysis, accommodating independent and various twin designs.

Main Results:

• BLVCM demonstrated superior performance in detecting interaction effects compared to existing methods (Granvil, SKAT) in simulations.
• Applied BLVCM to a twin study of over 20,000 gene regions, identifying significant RVs associated with low-density lipoprotein cholesterol levels.
• Identified novel gene regions exhibiting significant single nucleotide polymorphism-single nucleotide polymorphism (SNP-SNP) synergistic effects.

Conclusions:

• The BLVCM is an effective statistical tool for analyzing rare variant interactions in complex trait studies.
• The method successfully identified significant RVs and novel gene regions with synergistic effects in a real-world twin study.
• BLVCM offers a robust approach for gene-level and individual SNP contribution estimation, advancing the field of genetic association analysis.