Prediction of Health and Environmental Hazards of Chemical: A Hierarchical Approach Using QMSA and QSAR (1997-1998)

Abstract

During the past few years we have been involved in the development of new computational methods forquantifying similarity/dissimilarity of chemicals and applications of quantitative molecular similarity analysis (QMSA) techniques in analog selection and property estimation for use in the hazard assessment of chemicals. We have also explored the mathematical nature of the molecular similarity space in order to better understand the basis of analog selection by QMSA methods. The parameter spaces used for QMSA and analog selection were constructed from nonempirical parameters derived from computational chemical graph theory. Occasionally, graph invariants were supplemented with geometrical parameters and quantum chemical indices to study the relative effectiveness of graph invariants vis-a-vis geometrical and quantum chemical parameters in analog selection and property estimation. We carried out comparative studies of nonempirical descriptor spaces and physicochemical property spaces in selecting analogs. Molecular similarity methods were applied in predicting modes of toxic action (MOA) of chemicals. Our similarity/dissimilarity methods have also found successful applications in the discovery of new drag leads by US drag companies. In this project, we will have four primary goals: 1) development of a hierarchical approach to molecular similarity, 2) formulation of quantitative structure-activity relationship (QSAR) models for predictive toxicology using a hierarchical approach, 3) applications of hierarchical QSAR and QMSA approaches in computational toxicology related to human health and ecological hazard assessment, and 4) the application of hierarchical QMSA and QSAR approaches in estimating potential toxicity of deicing agents. The first goal of the project is the use of parameters of gradually increasing complexity, viz., topological, topochemical, geometrical, and quantum chemical indices, in the quantification of molecular similarity/dissimilarity of chemicals. We will take a two-tier approach in this area. First, similarity methods will be used in ordering sets of molecules and in selecting structural analogs of toxic chemicals which pose human health and ecological hazards. Secondly, we will use the properties of selected analogs in estimating toxicologically important properties for chemicals. Although different classes of parameters have been used in the characterization of molecular similarity, no systematic study has been carried out in the use of all four classes of parameters, mentioned above, in analog selection and property estimation. We will apply a hierarchical approach to the use of these four types of theoretical molecular descriptors in the quantification of molecular similarity/dissimilarity. The second goal consists of the development of hierarchical QSAR models for predicting the toxic potential of chemicals using topological and quantum chemical indices. Initially, we will use parameters calculated by semi-empirical methods such as MOP AC and AMP AC. Parameters calculated by ab initio quantum chemical methods will be used in limited cases of QSAR model development, if they are considered necessary. The third goal of the project will be the prediction of human health hazard and ecotoxicological effects of chemicals using QSAR and QMSA methods developed in the project. Attempts will be made to estimate endpoints, such as, carcinogenicity, mutagenicity, xenoestrogenicity, acute toxicity, transport of chemicals through the blood-brain barrier, biodegradation, and bioconcentration factor. The fourth goal will involve the utilization of QMSA and QSAR methods developed as part of this project in predicting the potential toxicity of deicing agents.