SBAT

SBAT (Sparse Bayesian Association Testing)

A program for performing Genome-wide association testing with multiple, continuous traits in related samples, using a multi-trait mixed model framework.

SBAT implements 2 different models

  1. A full model – where a SNP has a non-zero effect on all the traits.
  2. A sparse model – where a SNP has a non-zero effect on a subset of the traits.

SBAT is multi-threaded. It uses a combination of the Eigen matrix library and OpenMP to parallelise a number of linear algebra operations and also to analyse multiple SNPs concurrently.

Both models use an EM algorithm for genetic and environmental covariance estimation as the first step. This provides very efficient inference that allows hundreds of traits to be analysed. The computational efficiency of this step is illustrated in the Figure below.

timing

Software registration and license:

SBAT is freely available for academic use only.

For non-academic use see the LICENCE file (also included with each software download). Please register for access to the software here.

If you have any questions about this software please post a message on the OXSTATGEN mail list.

Citations

If you use full model in SBAT please cite the following publications:

Elliott et al. (2018) Genome-wide association studies of brain imaging phenotypes in UK Biobank. Nature 562pages 210–216.

Andrew Dahl, Valentina Iotchkova, Amelie Baud, Åsa Johansson, Ulf Gyllensten, Nicole Soranzo, Richard Mott, Andreas Kranis, Jonathan Marchini. A multiple phenotype imputation method for genetic studies. Nature Genetics doi:10.1038/ng.3513

If you use the sparse model in SBAT please cite

Kevin Sharp, Valentina Iotchkova and Jonathan Marchini (2017) Sparse Bayesian Modelling for Multitrait Genetic Association Studies (in preparation)