In silico scaling and prioritization of chemical disposition and chemical toxicity of 15,145 organic chemicals

Authors: Matthews EJ
Publication: Computational Toxicology
Software: ADMET Predictor®


This report describes the development and beta-test of methods that prioritize and scale in silico predictions of chemical disposition {(CD) intestinal absorption, membrane permeability, distribution, sequestration, toxicokinetics} and chemical toxicity {(CT) genetic, carcinogenicity, developmental, teratology}. It reports > 4.5 million data records for 15,145 organic chemicals that are ingredients in 32 different purpose groups of foods, drugs, and cosmetics, and 3682 colorants assigned to 36 chemical classes. Prioritization and scaling were facilitated by using: reference databases with known activities; data pre-processing; and prioritization P-metrics and P-indices to accentuate prediction strength of evidence and discriminate negative, indeterminate, and positive predictions. CD is computed based upon molecular descriptor properties of 1860 reference pharmaceuticals. CT is based upon predicted toxicities of 11,810 reference bioavailable test chemicals predicted with 3 software programs, 130 QSAR or expert system models. Results reveal major differences in toxicological activities: flavors and fragrances had very low predicted CD/CT activities, conversely, many colorants had high predicted genetic and carcinogenic activities. Contrasting CD/CT activities of flavors and hair dyes are presented. Furthermore, colorant chemical classes had different toxicological profiles, and a few classes having safer profiles were identified. Prediction data are also evaluated for DNA-reactivity as specified in ICH M7, persistent organic chemicals, and an alternative testing scheme is suggested to incorporate this methodology.