ABSTRACT
The purpose of global gene expression assays is to help to explain phenotypes such as disease manifestations, drug response, or diet effect. The commonly used methods of statistical analysis allow one to establish a correlation between the expression prole (usually presented as a selected “gene signature”) and the phenotype (reviewed in Draghici [1]), but they are insufcient for understanding the mechanism behind the phenotype. Moreover, gene signatures as descriptors are notorious for poor robustness and reproducibility across platform and experimental conditions. It is now widely accepted that the methods of functional, “systems level” analysis such as pathways, networks, and “interactome” modules are needed for interpretation of “Omics” data sets [2]. For years, functional tools such as KEGG maps [3] and GO ontology [4] were mostly used as a follow-up step after statistical analysis as an attempt to understand the underlying biology behind the gene signature. Since most statistics-derived gene lists make little functional sense, the potential of functional analysis in Omics data mining stayed largely unrealized.
