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

摘要

This project focuses on a two-pronged approach to modeling toxicity data. A standard structure-based QSAR approach using chemodescriptors (descriptors based on the chemical structure of the toxicant) has been coupled with the development of biodescriptors, a novel set of mathematical descriptors derived from 2-DE proteomic gel analyses. The research group has explored the use of chemodescriptors calculated using high-level chemodescriptors as well as considering potential mechanistic approaches to chemical toxicity, e.g., using vertical electron affinity for modeling the interactions of highly reactive chemicals. Biodescriptor development has focused on three main approaches: global descriptors that characterize the entire proteomics map, local descriptors that characterize a subset of the proteins present in the gel, and spectrum-like descriptors. These efforts have been further enhanced through the use of robust statistical approaches and the development of new statistical techniques for analyzing the full set of proteins present in a proteomics map.