The SPARROW method uses a sparse regression methodology, variational Bayes spike regression (VBSR), to infer the relative importance of a given candidate expression driver. This is done by fitting a sparse regression model for every gene in an expression data-set with a set of candidate expression drivers as potential drivers. Candidate drivers which are frequently chosen in the sparse bases across genes are prioritized as more likely to be true gene expression drivers. An R implementation of the VBSR algorithm is freely available from CRAN: http://cran.r-project.org/web/packages/vbsr/index.html (academic use only). Here is a vignette showing the basic functionality of the VBSR package: [pdf] |