BioMark is an R package implementing two meta-strategies for finding biomarkers. For two-class discrimination problems, Higher Criticism is available, an approach to define optimal cutoffs for methods like t testing for high-dimensional data. In this work, HC has been extended to multivariate methods, allowing, e.g., selection of biomarkers based on the size of regression coefficients. A second approach is Stability Selection, considering the consistency of variable selection under perturbation of the data. Both methods have the advantage that they do not rely on error estimates, that in this setting tend to be highly variable.
Wehrens, H.R.M.J.; Franceschi, P. (2013). Biomarker selection in R: the BioMark package. In: High-Throughput Omics & Data Integration Workshop (COST Action SeqAhead), Barcelona, February 12-15, 2013. handle: http://hdl.handle.net/10449/23463