| Regulatory toxicology has traditionally relied on studies in laboratory animals to characterize the potential hazards of chemicals and commercial products for human health effects. Not only do these animal models carry uncertainties in extrapolating to human safety evaluations, but they also suffer from low-throughput, are time and money consuming, and have inherent ethical concerns. These issues have heightened the need for alternative approaches that can better predict and evaluate the effects that chemicals can have on biological processes. In this thesis, a comparative quantitative structure-activity relationship (QSAR) study was undertaken and derived from the mutagenic potency of 88 mutagenic aromatic amines (AAs) acting on Salmonella typhimurium TA 98 + S9 and 67 mutagenic AAs acting on TA 100 + S9 reported by Debnath (1992a). Using a stepwise multiple linear regression algorithm, novel QSAR models integrating a structural indicator variable (IL), Log P, and molar refractivity (MR) (new Ames TA 98 QSAR) and Log P and chemical hardness () (new Ames TA 100 QSAR) were developed and statistically compared to the respective Debnath models. Internal predictive performance of the models was analyzed and compared via goodness of fit statistics and leave-one-out cross validation. The performance of external prediction was evaluated using a test set of AAs that were not included in the training set used to develop the model. It was found that both new QSAR models, while utilizing one less descriptor variable, exhibited greater goodness of fit statistics and internal performance than the respective Debnath models. However, the new QSAR models did not fair as well (albeit comparable) as Debnath models in predicting the external test set compounds. Structural features of the poorly predicted AAs from both the internal and external validations were compared and important insights into the mechanism of toxic action for AAs are discussed. With transparent communication of QSARs, these non-testing in silico methods have the potential to screen and prioritize large numbers of chemicals and usher in new opportunities to anticipate undesirable toxicological effects of chemicals. |
