The goal of mrbma is to implement Mendelian Randomization Bayesian Model Averaging. This is a form of multivariable Mendelian randomization which can be useful for prioritizing likely-causal risk factors among a set of highly-correlated exposures. This code adapts the approach described in:
Zuber V, Colijn JM, Klaver C, Burgess S. Selecting likely causal risk factors from high-throughput experiments using multivariable Mendelian randomization. Nature Communications. 2020 Dec 1;11(1):1–11.
This code further implements an empirical permutation procedure to derive p-values using the Nyholt procedure of effective tests to control the false discovery rate. This approach was utilized in:
Levin MG, Zuber V, Walker VM, Klarin D, Lynch J, Malik R, et al. Prioritizing the Role of Major Lipoproteins and Subfractions as Risk Factors for Peripheral Artery Disease. Circulation. 2021 Aug 3;144(5):353–64.
You can install the development version of mrbma from GitHub with:
# install.packages("devtools")
devtools::install_github("mglev1n/mrbma")