invt.io/1lxb4q5tgye

Make easy use of our method by using our R-package. Find instructions for installation and a guide to getting started on the official documentation.
GitHub: github.com/lcrawlab/sme.
CRAN: cran.r-project.org/package=smer.
Documentation: lcrawlab.github.io/sme.
6/6

In real data analysis of complex traits of the UK Biobank, we illustrate that SME, informed by DNase-seq data, identifies statistical epistasis of variants for which previous research have also found interaction pathways. 5/6

The sparsity in SME leads to a significant increase in power, while controlling type I error. Additionally, SME leverages a novel algorithm to substantially improve its scalability for genome-wide analyses. SME runs 10 to 90 times faster than state-of-the-art epistatic mapping methods! 4/6

In SME, we incorporate trait specific information on functional enrichment into linear mixed models to induce sparsity in the modeled gene interactions. This improves both the efficiency of statistical estimators and scalability of the test for marginal epistasis. 3/6

Epistatic contributions to the variance of polygenic traits are notoriously hard to detect. By focusing on marginal epistasis, the combined pairwise interaction effects between a given variant and all other variants, we overcome small effect sizes and reduce the search space. 2/6

The sparsity in SME leads to a significant increase in power, while controlling type I error. Additionally, SME leverages a novel algorithm to substantially improve its scalability for genome-wide analyses. SME runs 10 to 90 times faster than state-of-the-art epistatic mapping methods! 4/6

In SME, we incorporate trait specific information on functional enrichment into linear mixed models to induce sparsity in the modeled gene interactions. This improves both the efficiency of statistical estimators and scalability of the test for marginal epistasis. 3/6

Epistatic contributions to the variance of polygenic traits are notoriously hard to detect. By focusing on marginal epistasis, the combined pairwise interaction effects between a given variant and all other variants, we overcome small effect sizes and reduce the search space. 2/6