The Sparse Marginal Epistasis test

The smer package implements a computationally and statistically efficient method for detecting marginal epistasis in genome-wide association studies (GWAS). Find the full package documentation including examples and articles here: Sparse Marginal Epistasis test Documentation.

Key Features

• Hutchinson’s stochastic trace estimator: efficient and scalable computation
• Mailman algorithm: fast vector-by-matrix operation
• Linear mixed model: controls for population structure
• Multimodal Input: incorporates additional data from HDF5 files to improve power in detecting gene-by-gene interactions.
• Optimize for Memory Constraints: Highly configurable block wise processing of the data allows to make the most of available resources. See also How To Optimize the Memory Requirements of SME.
• Parallelization: Utilizes OpenMP for multi-threaded processing.

Installation

Installation from CRAN

You can install the latest release from CRAN

install.packages("smer")

Installation from source

You can install the development version of smer from GitHub with:

install.packages("devtools")
devtools::install_github("lcrawlab/sme")

Dependencies

System requirements of the package:

• GNU make
• R (>= 4.4)
• Rhdf5lib (from BioConductor)
• OpenMP (optional)

To install Rhdf5lib, first install the tool BiocManager from CRAN, then install the library using this tool.

if (!require("BiocManager", quietly = TRUE))
    install.packages("BiocManager")
BiocManager::install("Rhdf5lib")

The full list of R dependencies can be found in the DESCRIPTION file.

OpenMP

For OS X and Linux, the OpenMP library can be installed via one of the (shell) commands specified below:

System	Command
OS X (using Homebrew)	brew install libomp
Debian-based systems (including Ubuntu)	sudo apt-get install libomp-dev

To enable openMP, it may be necessary to configure the compiler flags SHLIB_OPENMP_CXXFLAGS and LDFLAGS in the ~/.R/Makevars file.

System	Required Flags
OS X	-Xclang -fopenmp -lomp
Linux	-fopenmp -lomp

Known Issues

If the error is ld: library "crypto" not found, install openssl (e.g. brew install openssl).

Compiling the package requires the compiler to find the libraries for the dependencies. For unix systems, the libraries are typically installed at /usr/local/lib and /usr/local/include. For users using OS X and homebrew, the libraries are typically installed at /opt/homebrew/lib and /opt/homebrew/include.

Non-standard library paths need to be configured. The src/Makevars file configures the compiler flags and considers the LDFLAGS and CPPFLAGS from the ~/.R/Makevars file.

References

• Stamp J, Crawford L (2025). smer: The Sparse Marginal Epistasis Test. R package version 0.0.1, https://lcrawlab.github.io/sme/, https://github.com/lcrawlab/sme.
• Stamp J, Smith Pattillo S, Weinreich D, Crawford L (2025). Sparse modeling of interactions enables fast detection of genome-wide epistasis in biobank-scale studies. biorxiv, https://doi.org/10.1101/2025.01.11.632557
• Stamp J, Crawford L (2024). mvMAPIT: Multivariate Genome Wide Marginal Epistasis Test. R package version 2.0.3, https://lcrawlab.github.io/mvMAPIT/, https://github.com/lcrawlab/mvMAPIT.
• Stamp et al. (2023): Leveraging genetic correlation between traits for epistasis detection in GWAS. G3: Genes, Genomes, Genetics.
• Fu, B., Pazokitoroudi, A., Xue, A., Anand, A., Anand, P., Zaitlen, N., & Sankararaman, S. (2023). A biobank-scale test of marginal epistasis reveals genome-wide signals of polygenic epistasis. bioRxiv.
• Crawford et al. (2017): Detecting epistasis with the marginal epistasis test. PLoS Genetics.
• Devresse et al. (2024): HighFive - Header-only C++ HDF5 interface. https://zenodo.org/records/13120799