BayesXII is a fast parallelized algorithm for Bayesian regression models, where "X" stands for Bayesian alphabet methods and "II" stands for "parallel". Ideally, the computations at each step of the MCMC chain can be accelerated by k times, where k is the number of computer processors, up to p times, where p is the number of markers.
Details can be found in this paper:
Zhao T, Fernando R, Garrick D, et al. Fast parallelized sampling of Bayesian regression models for whole-genome prediction[J]. Genetics Selection Evolution, 2020, 52(1): 1-11.
All codes are provided here, and a demo using a small dataset is provided. The parallel implementation of the BayesXII algorithm was based on Message Passing Interface (MPI), and the tutotial of using MPI in Julia is provided in my repository named MPI_testing.
Step 1. run create_data_demo.jl to create data such as genotype and phenotype. Output data are in total about 3 MB, which are also provided in the demo folder.
Step 2. run split_column.jl to create data specifically needed by different processors. Output data are in total 2 MB, which are also provided in the demo folder.
Step 3. run the MPI parallel code BayesXII.jl on multiple processors. A slurm submit file is provided in the demo folder to submit parallel jobs.
Feel free to contact Hao Cheng [qtlcheng@ucdavis.edu] and Tianjing Zhao [tjzhao@ucdavis.edu] for any problem.