The bmass R package provides accessible functions for running the algorithms described in Stephens 2013 PLoS ONE and applied to multiple large, publicly available GWAS datasets in Turchin and Stephens 2019. bmass conducts a Bayesian multivariate analysis of GWAS data using univariate association summary statistics. Output inclues whether any new SNPs are found as multivariate genome-wide significant as well as posterior probabilities of each significant SNP’s assignment to different multivariate models.
For more details on the results of applying bmass to publicly available GWAS datasets, please see our paper in PLOS Genetics. For more details regarding the underlying algorthims of bmass, please see Stephens 2013 PLoS ONE.
If you find a bug, or you have a question or feedback on our work, please post an issue.
Citing this work
If you find the bmass package or any of the source code in this repository useful for your work, please cite:
Turchin MC and Stephens M (2019) “Bayesian multivariate reanalysis of large genetic studies identifies many new associations.” PLOS Genetics 15(10): e1008431. doi.org/10.1371/journal.pgen.1008431
License
Copyright (c) 2016-2019, Michael Turchin and Matthew Stephens.
All source code and software in this repository are made available under the terms of the MIT license. See file LICENSE for the full text of the license.
Quick Start
To install bmass from CRAN:
install.packages("bmass")To install the most recent dev version of bmass from github:
install.packages("devtools")
devtools::install_github("mturchin20/bmass@v1.0.3", build_vignettes=TRUE)Once you have installed the package, load the package in R:
library("bmass")Next, view and run the example code provided in the first introductory vignette using simulated data. A second, more advanced introductory vignette is also available that involves downloading, processing, and analyzing the GlobalLipids 2013 data.
Credits
The bmass R package was developed by Michael Turchin at the University of Chicago, with contributions from Peter Carbonetto and Matthew Stephens, and based on the R code provided in Stephens 2013 PLoS ONE.
