mvMAPIT

mvMAPIT detects marginal epistasis across multiple correlated traits using a multivariate linear mixed-model framework to leverage genetic correlations and identify variants involved in pairwise interaction effects.

Key Features:

• Multivariate Approach: Incorporates multiple outcomes or traits simultaneously to leverage genetic correlation between traits for improved identification of variants involved in epistasis.
• Marginal Epistasis Detection: Focuses on detecting marginal epistasis by estimating combined pairwise interaction effects between a given variant and all other variants without pinpointing specific interacting partners.
• Linear Mixed Model Formulation: Formulated as a multivariate linear mixed model with parameter inference and P-value computation via a multitrait variance component estimation algorithm.
• Scalability: Designed to be scalable for moderately sized GWAS and applicable to large-scale genetic studies.
• Simulation Validation: Simulations demonstrate advantages over univariate epistatic mapping strategies, including improved statistical power and efficiency.

Scientific Applications:

• Protein sequence analysis of broadly neutralizing anti-influenza antibodies: Applied to protein sequence data from broadly neutralizing anti-influenza antibodies to identify variants involved in epistatic effects.
• Heterogeneous stock mice GWAS: Applied to approximately 2,000 heterogeneous stock mice from the Wellcome Trust Centre for Human Genetics to detect epistatic variant signals across traits.

Methodology:

mvMAPIT uses a multivariate linear mixed model that captures trait correlation structure and employs a multitrait variance component estimation algorithm for parameter inference and P-value computation while focusing on marginal epistasis (combined pairwise interaction effects between a variant and all other variants).