mfsusieR

The mfsusieR package implements multi-outcome functional regression using the Sum of Single Effects (mfSuSiE) model. The method is general for Bayesian variable selection in functional-regression problems, but was motivated and developed for fine-mapping spatially correlated genomic features (e.g., DNA methylation profiles, chromatin accessibility tracks) where the outcome is a curve along the genome and a sparse set of variables affects several correlated positions jointly. Built on the susieR backbone via S3 dispatch.

The package exposes two main functions:

• fsusie(Y, X, pos = NULL, ...) for fine-mapping a single response. Y may be scalar or functional (a matrix sampled at a grid of positions).
• mfsusie(X, Y, pos = NULL, ...) for fine-mapping multiple responses jointly. Y is a list of length M outcomes; each element is a matrix n x T_m (with T_m = 1 for scalar outcomes and T_m > 1 for functional outcomes; T_m may differ across outcomes).

When Y is scalar, fsusie() reduces to a version of the SuSiE model. We expose this case here for two reasons: (i) it provides a sanity-check path against susieR (the C1 contract test suite locks exact element-wise equivalence), and (ii) it allows scalar outcomes to be fit jointly with functional outcomes through mfsusie(). Users analyzing a single scalar response on its own should use the susieR package directly. Users with multiple correlated scalar outcomes without a spatial structure (e.g., several QTL traits across tissues) should use mvsusieR instead. mfsusieR is the right choice when one or more outcomes are functional, or when functional and scalar outcomes are jointly modelled.

Installation

# install.packages("remotes")
remotes::install_github("stephenslab/susieR")
remotes::install_github("StatFunGen/mfsusieR")

mfsusieR currently depends on the GitHub master of susieR for the per-iteration S3 generics. CRAN and conda releases of both packages are planned; once those land the dependency on the GitHub master will be dropped and a single install.packages() (or conda install) will be sufficient.

Quick start

See the pkgdown website, in particular the Getting Started vignette, for worked examples in both single- and multi-outcome contexts.

Citing this work

If you use mfsusieR::fsusie() (single-outcome functional fine-mapping) in your work, please cite:

Denault, W.R.P., Sun, H., Carbonetto, P., Liu, A., De Jager, L.P., Bennett, D., The Alzheimer’s Disease Functional Genomics Consortium, Wang, G. & Stephens, M. (2025). fSuSiE enables fine-mapping of QTLs from genome-scale molecular profiles. bioRxiv DOI: 10.1101/2025.08.17.670732

If you use mfsusieR::mfsusie() (multi-outcome joint fine-mapping) in your work, please cite:

Liu, A., Sun, H., De Jager, L.P., Bennett, D., The Alzheimer’s Disease Functional Genomics Consortium, Wang, G. & Denault, W.R.P. (2025). mfSuSiE enables multi-cell-type fine-mapping and multi-omic integration of chromatin accessibility QTLs in aging brain. bioRxiv DOI: 10.1101/2025.11.25.690439

For the underlying SuSiE backbone and engineering improvements that mfsusieR builds on, please also cite:

McCreight, A., Cho, Y., Li, R., Nachun, D., Gan, H-Y., Carbonetto, P., Stephens, M., Denault, W.R.P. & Wang, G. (2025). SuSiE 2.0: improved methods and implementations for genetic fine-mapping and phenotype prediction. bioRxiv DOI: 10.1101/2025.11.25.690514

License

BSD 3-Clause. See LICENSE.

Issues and contributions

Please file an issue for bug reports, questions, or suggestions.