Computational methods to predict drug safety liabilities.

Computational methods to predict drug safety liabilities are reviewed. A special emphasis of this article is on the perceived strengths and weaknesses of the commercial turnkey predictive toxicology programs (TOPKAT, MULTICASE and DEREK). This article includes proposals for improvements of individual predictive programs, experiences with pharmaceutical datasets, evaluations incorporating multiple programs, and strategies of their use as a sentinel filter for liability assessment early in the drug discovery process.