In silico approaches to predicting cancer potency for risk assessment of genotoxic impurities in drug substances.

The current risk assessment approach for addressing the safety of very small concentrations of genotoxic impurities (GTIs) in drug substances is the threshold of toxicological concern (TTC). The TTC is based on several conservative assumptions because of the uncertainty associated with deriving an excess cancer risk when no carcinogenicity data are available for the impurity. It is a default approach derived from a distribution of carcinogens and does not take into account the properties of a specific chemical. The purpose of the study was to use in silico tools to predict the cancer potency (TD(50)) of a compound based on its structure. Structure activity relationship (SAR) models (classification/regression) were developed from the carcinogenicity potency database using MultiCASE and VISDOM. The MultiCASE classification models allowed the prediction of carcinogenic potency class, while the VISDOM regression models predicted a numerical TD(50). A step-wise approach is proposed to calculate predicted numerical TD(50) values for compounds categorized as not potent. This approach for non-potent compounds can be used to establish safe levels greater than the TTC for GTIs in a drug substance.

How well do lipophilicity parameters, MEEKC microemulsion capacity factor, and plasma protein binding predict CNS tissue binding?

Brain fraction unbound (Fu) is critical to understanding the pharmacokinetics/dynamics of central nervous system (CNS) drugs, thus several surrogate predictors have been proposed. At present, correlations between brain Fu, microemulsion electrokinetic chromatography capacity factor (MEEKC k'), plasma Fu, octanol-water partition coefficient (clogP), and LogP at pH 7.4 (clogD(7.4) ) were compared for 94 diverse molecules, and additionally for 587 compounds. MEEKC k' was a better predictor of brain Fu (r(2) = 0.74) than calculated lipophilicity parameters (clogP r(2) = 0.51-0.54, clogD(7.4) r(2) = 0.41-0.44), but it was not superior to plasma Fu (r(2) = 0.74-0.85) as a predictor of brain Fu. MEEKC k' did not predict plasma Fu(r(2) = 0.58) as well as brain Fu, and the extent of improvement over clogP or clogD(7.4) (r(2) = 0.41-0.49) was less pronounced. Although log-log-correlation analysis supported seemingly strong prediction of brain Fu both by MEEKC k' and by plasma Fu (r(2) ≥ 0.74), analysis of prediction error estimated a 10-fold and 6.9-8.6-fold prediction interval for brain Fu estimated using MEEKC k' and plasma Fu, respectively. Therefore, MEEKC k' and plasma Fu can predict the log order of CNS tissue binding, but they cannot provide truly quantitative brain Fu predictions necessary to support in-vitro-to-in-vivo extrapolations and pharmacokinetic/dynamic data interpretation.