Repository files navigation

Thesmer package implements a computationally and statisticallyefficient method for detecting marginal epistasis in genome-wideassociation studies (GWAS). Find the full package documentationincluding examples and articles here:Sparse Marginal Epistasis testDocumentation.

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

You can install the latest release from CRAN

install.packages("smer")

You can install the development version ofsmer fromGitHub with:

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

System requirements of the package:

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

To installRhdf5lib, first install the toolBiocManager 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 theDESCRIPTIONfile.

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

To enable openMP, it may be necessary to configure the compiler flagsSHLIB_OPENMP_CXXFLAGS andLDFLAGS in the~/.R/Makevars file.

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

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

Non-standard library paths need to be configured. Thesrc/Makevarsfile configures the compiler flags and considers theLDFLAGS andCPPFLAGS from the~/.R/Makevars file.

• Stamp J, Crawford L (2025). smer: The Sparse Marginal Epistasis Test. Rpackage version 0.0.1,https://lcrawlab.github.io/sme/,https://github.com/lcrawlab/sme.
• Stamp J, Smith Pattillo S, Weinreich D, Crawford L (2025). Sparsemodeling of interactions enables fast detection of genome-wideepistasis in biobank-scale studies. biorxiv,https://doi.org/10.1101/2025.01.11.632557
• Stamp J, Crawford L (2024). mvMAPIT: Multivariate Genome Wide MarginalEpistasis 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 forepistasis 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 epistasisreveals genome-wide signals of polygenic epistasis. bioRxiv.
• Crawford et al. (2017): Detecting epistasis with the marginalepistasis test. PLoS Genetics.
• Devresse et al. (2024): HighFive - Header-only C++ HDF5 interface.https://zenodo.org/records/13120799