Skip to content

mvSuSiE

Summary

mvSuSiE is an open-source R package (mvsusieR) for multitrait statistical fine-mapping. It generalizes the single-trait Sum of Single Effects (SuSiE) model to joint multiple regression, allowing researchers to fine-map multiple related traits simultaneously. By learning patterns of shared genetic effects directly from data, mvSuSiE increases power to identify causal variants, achieves narrower credible sets, and distinguishes whether causal signals are trait-specific or shared.

Method and Capabilities

mvSuSiE introduces several key advances for multitrait analysis: - Shared Effect Modeling: It utilizes a flexible prior (a mixture of multivariate normal distributions) to model heterogeneous effect-sharing patterns across traits, estimated using Extreme Deconvolution (ED). - Trait-Wise Significance: To determine which traits a causal variant affects, mvSuSiE computes the Local False Sign Rate (lfsr) for each variant-trait pair. - Summary Statistics Support: Supports fine-mapping using individual-level data or summary statistics (z-scores and in-sample LD).

mvSuSiE has been applied to jointly fine-map 16 blood cell traits in the UK Biobank, characterizing complex pleiotropic architectures.

Citations

  • Zou, Y., Carbonetto, P., Xie, D., Wang, G., & Stephens, M. (2026). Fast and flexible joint fine-mapping of multiple traits via the Sum of Single Effects model. Nature Genetics, 58(2), 454–462. Source paper: s41588-025-02486-7.pdf