From QSAR to QNAR, developing enhanced models for drug discovery

Exploring new chemical entities in drug discovery requires extensive investigations on libraries of thousands of molecules. While conventional animal-based tests in drug discovery procedure are expensive and time consuming, the evaluation of a drug candidate can be facilitated by alternative computational methods. For example, the Quantitative Structure Activity Relationship (QSAR) model has been widely used to predict bioactivities for drug candidates. However, traditional QSAR models are solely based on chemical structures, and are less effective in the drug discovery procedure due to various limitations related to complicated structures or bioactivities. In this thesis, we aimed to establish high quality and predictive models by using novel modeling approaches beyond QSAR. First, we developed a methodology for predicting the Blood-Brain Barrier permeability of small molecules by incorporating biological assay information (e.g. transporter interactions) into the modeling process. This method can be further extended to modeling and predicting in vivo bioactivities of drug candidates. Second, we created a new Quantitative Nanostructure Activity Relationship (QNAR) modeling strategy to extend the applicability of QSAR to predict bioactivities of nanomaterials. The research presented in this thesis opens a new path to the precise prediction of bioactivities of molecules in the drug discovery procedure.Ph.D.Includes bibliographical referencesby Wenyi Wan