Use of Predictive Toxicology in the Design of New Chemicals
Vijay K. Gombar, Kurt Enslein · ACS symposium series · 1995
While many applications of computer-assisted techniques involve the design of molecules with maximum biological activity, this chapter addresses a different criterion of molecular design. We consider monitoring toxicity as the design process brings structural changes because the most desirable chemical is not necessarily the one with maximum activity but rather the one with maximum activity-to-toxicity ratio. The computer-assisted technique for predicting toxicity profiles described here relies on robust and crossvalidated quantitative structure-toxicity relationship (QSTR) models developed from experimentally determined animal toxicity data. The technique, as embedded in the software package TOPKAT, first confirms whether the query chemical lies inside, near the boundary of, or outside the optimum prediction space (OPS) of a QSTR model in order to assess the reliability of predictions. Currently, such toxicity measures as rodent carcinogenicity, mutagenicity in Salmonella typhimurium teratogenicity, chronic lowest observed adverse effect level (LOAEL) and maximum tolerated dose (MTD), acute rat oral LD 50 and mouse inhalation LC 50, and aquatic EC 50 and LC 50 can be estimated by using these techniques. The methodology is explained here using developmental toxicity (DT) data.