RESUMO
Originally conceptualized as an integrated approach combining conventional toxicology methods with genome-wide expression profiling, toxicogenomics has promised to provide unequivocal relationships between the molecular changes elicited by a compound or a target pathway and the lesions that appear subsequently in the tissues. However, the discipline has only partially delivered on this promise, and the number of publications and submissions related to toxicogenomics is stagnating. The purpose of this article is to outline key factors contributing to a successful implementation of toxicogenomics in the drug discovery and development process. Paradigms and methods of toxicogenomics are briefly reviewed, and the prominence of biostatistics and its limitations in the particular context of nonclinical toxicology studies are discussed. We present specific approaches for pathophysiological contextualization of gene expression data derived from tissues with lesions at variable incidence and severity: "unmixing" (deconvolution) of molecular expression profiles from complex tissues, the invaluable contribution of reference data, the role of establishing causation between expression signals and pathologic changes (phenotypic anchoring), and especially molecular localization. These approaches compensate for the limitations of biostatistical analysis, which in turn, derive from tissue heterogeneity. Finally, impactful applications of toxicogenomics along the drug discovery and development process are exemplified, from the evaluation of potential target toxicities to the selection of candidate compounds and elucidation of the molecular and cellular mechanisms leading to chronic toxicity.