Multi-gene expression-based statistical approaches to predicting patients' clinical outcomes and responses.

Gene expression profiling technique now enables scientists to obtain a genome-wide picture of cellular functions on various human disease mechanisms which has also proven to be extremely valuable in forecasting patients' prognosis and therapeutic responses. A wide range of multivariate techniques have been employed in biomedical applications on such expression profiling data in order to identify expression biomarkers that are highly associated with patients' clinical outcome and to train multi-gene prediction models that can forecast various human disease outcome and drug toxicities. We provide here a brief overview on some of these approaches, succinctly summarizing relevant basic concepts, statistical algorithms, and several practical applications. We also introduce our recent in vitro molecular expression-based algorithm, the so-called COXEN technique, which uses specialized gene profile signatures as a Rosetta Stone for translating the information between two different biological systems or populations.