Quantitative methods to predict bio-potency and bioavailability

Free energy force field three-dimensional quantitative structure-activity relationship (FEFF 3D-QSAR) analysis has been applied to a series of gadolinium chelating compounds that bind to human serum albumin (HSA) and serve as magnetic resonance imaging (MRI) contrast agents. QSAR models constructed using FEFF 3D-QSAR analysis reveal that the desolvation energy, change of electrostatic binding energy and total van der Waals binding energy have dominant effects on the binding profile of the training set of ligands. Important geometric factors governing binding can also be inferred from the equilibrium ligand-receptor complex structure. Overall, a small number of the thermodynamic contributions can be used as descriptors in the QSAR equation to effectively predict ligand binding affinity.; Membrane-interaction (MI) 3D-QSAR analysis is an extended and specialized version of FEFF 3D-QSAR which allows a user to (1) construct a model membrane monolayer as “general receptor” in a receptor-dependent (RD) 3D-QSAR paradigm, (2) include user selected non-FEFF 3D-QSAR type descriptors in the trial QSAR descriptor pool, (3) use any target biological endpoint as the dependent variable when constructing QSAR models.; MI-3D-QSAR models have been constructed from a training set containing thirty structurally diverse molecules whose permeability coefficients across the cellular membranes of Caco-2 cells have been measured. The models reveal the two major factors that affect the permeability of a molecule are the solubility of the solute and solute-membrane binding. Eight molecules have been used as a test set to check the predictivity of the MI-QSAR model. Predicted and observed permeability values agree with each other quite well (R > 0.9) for both the training set molecules and the test set molecules.; Based on the results reported in this thesis, FEFF 3D-QSAR and MI-QSAR are each powerful research tools in the area of molecular modeling.