Use of Biodescriptors and Chemodescriptors in Predictive Toxicology: A Mathematical/Computational Approach

Abstract

Development and applications of biodescriptors for predictive toxicology by our NRRI team has been expanded to incorporate three major types of biodescriptors: a) global descriptors from invariants of matrices associated with proteomics maps, b) a set of local invariants describing various aspects of each map (instead of one global biodescriptor), and c) spectrum-like descriptors for the characterization of proteomics patterns.A successful HiQSAR model has been developed for a set of 55 halocarbons for which cellular level toxicity data is available. It may be noted that this is the "superset" of compounds from which the "subset" of twenty halocarbons, currently being tested by WPAFB and Dr. Witzmann using the DNA microarray and proteomics analysis, was selected. Selection of parameters for the HiQSAR was based on the mechanistic hypothesis that dissociate electron attachment and subsequent formation of free radical$, leading to lipid peroxidation, is a major factor in halocarbon toxicity. This conclusion was derived from Dr. Balasubramanian's previous research, based on high-level quantum chemical calculations. If the hypothesis is correct, calculated parameters such as vertical electron aftinity(VEA) should be strongly related to a chemical's toxicity.