Development And Application Of Drug Design Methods Based On Binding Thermodynamics And Kinetics

The binding affinity, can be translated into a thermodynamic parameter, such as association constant(KA), dissociation constant(KD) or binding free energy(AGbinding), is the important parameter for characterizing the binding strength of a drug to its target and the stability of a drug-target complex. The binding kinetic parameters, such as association and dissociation rate constants(kon and koff), or activation free energies of association and dissociation(△G*on≠and AGoff≠), are conducted to determining the binding thermodynamic parameter(KA=kon/koff) and strongly concerned with drug efficiency and toxicity, etc. The traditional paradigm for drug discovery and development is driven by binding affinity, emphasizing that the binding affinity is the molecular basis of pharmacology-the higher the binding affinity, the better potency of a drug. However, recent studies have shown that drug efficiency is directly proportional to binding kinetics rather than binding affinity. Hence, much attention has been paid on binding kinetics, and the drug discovery strategy begins to be re-established by appreciating both drug-target binding kinetics and binding thermodynamics. However, to date, efficient computational methods for estimating binding kinetics are still not available. Moreover, although comparatively great progresses have been made during past decades, the current computational methods for predicting binding affinity still have limitations and shortages, such as low accuracy or time-consuming. In addition, the existing molecular docking methods are not applicable to some specific targets, and most docking programs are not able to deal well with the flexibility of proteins and crucial water molecules, and cannot sufficiently sample the conformational space, and so on.Aimed at solving above scientific problems existing in computer-aided drug design, the present dissertation mainly focuses on computing binding thermodynamic and kinetic parameters of drug-target to develop new drug design methods, and performs a series of studies by applying these methods in drug design. Chapter1is a brief review and some expansion of the basic theories of drug-target binding, and also points out the research advances and existing problems for the current computational methods of binding thermodynamics and kinetics. The methodology development and application of computational thermodynamics for drug-target binding are presented in chapters2to6. Chapter2introduces a statistical thermodynamic model that can quantitatively describe activator-potassium ion channel interaction and concentration-response relationship for a ligand to activate an ion channel. This model was developed based on the principles of statistical mechanics including the Ising model and lattice theory, and the model also considered the action fashions and cooperative effects of ligand-ion channel interaction. In chaper2, it was demonstrated that our model may excellently fit the concentration-response data for ztz240, a KCNQ2activator, to potentiate the ion channel. Remarkably, the theoretical model may accurately predict that the concentration-response curve of ztz240activating the wide-type/F137A mutant heterozygous channel is bi-sigmoid. This prediction is in good agreement with the experimental data determined in parallel in this chapter. These results lend credence to the assumptions on which the model is based and the model itself. Chapter3devotes to improving the accuracy of conformational energy evaluation for the conformational sampling method, named Cyndi, developed in our previous work. Based on it, a multi-empirical criteria based conformation sampling method (MECBM) was developed, which features higher accuracy and higher calculation speed over other conformation sampling methods and is suit for being used in large-scale drug design. By combing MECBM with other drug design methods, a virtual screening on two compound databases targeting the epidermal growth factor receptor (EGFR) T790M/L858mutant was performed. A series of candidates were verified as active inhibitors of EGFR T790M/L858mutant by experimental bioassays. Through a cycle of structural optimization, several low nanomol irreversible inhibitors of EGFR-T790M/L858were obtained. Chapter4covers the target specific scoring function design and docking program development for DNA and metalloproteins, respectively. By using the Non-dominated Sorting Genetic Algorithm-Ⅱ (NSGA Ⅱ), a multi-objective optimization model based docking program called iDNASBinder was developed to specifically simulate the ligand-DNA binding, and a multi-objective optimization model based docking program called iMpSDock was developed to specifically simulate the ligand-metalloprotein binding.iDNASBinder was verified by drug design practice. Combined with experimental bioassay,iDNASBinder was used to discover new inhibitors against the cognate target DNA sequence of transcription factor activator protein1(AP-1). As a result, the natural products veratramine and its two analogues were discovered as inhibitors of AP-1.iMpSDock can accurately predict various formation and reasonably quantify the thermodynamic properties of metal-ligand coordination bonds. To validate the reliability of iMpSDock in drug discovery, a docking simulation to predict the binding modes of quinoline-4-carboxylates to farnesyltransferase (FTase) was performed by using iMpSDock, which addressed the bidentate binding modes of quinoline-4-carboxylates to the zinc ions of FTase. On the basis of the biological result and predicted binding modes, an analysis on the structure-activity relationship of the quinoline-4-carboxylates and revealed some clues for modifying and optimizing of these compounds was performed. Chapter6deals with the methodology of binding free energy predication and flexible docking program. First of all, a binding free energy calculation method was developed by modifying MM-GBSA, an approach combines molecular mechanics, the generalized Born model and solvent accessibility method to calculate the binding free energy. The improved MM-GBSA method has reformed and reweighted the original energy terms. Using the improved MM-GBSA equation as the scoring function, a flexible docking program named iFitDock was developed. iFitDock has considered the flexibilities for both ligands and proteins, and can explicitly deal with the conserved water molecules embedded in the protein. iFitDocck can accurately predict the binding conformations for ligands to proteins, and can characterize the water molecules mediated binding interaction of drugs to their proteins. In this chapter, iFitDock was successfully applied to predict a new binding site for2-chloro-4-methyl-N-(naphthalene-2-yl)thiazole-5-carboxamide to the human dihydroorotate dehydrogenase (hDHODH), and the predicted result has been proven by the X-ray crystal structure determination of hDHODH in complex with this inhibitor. Chapter7is associated with the methodology development and application for the predication of both ligand-protien binding affinity and kinetics. Based on the improved MM-GBSA method as well as iFitDock developed in chapter6, and adopting the energy landscape theory that has been widely used in studying protein folding and the transition state theory that has been extensively employed in investigating chemical reactions, a method that can be used to properly construct the ligand-protein binding free energy landscape (BFEL) was developed. From the BFEL, lowest binding free energy pathway(s) for a ligand entering to the binding site of a protein can be addressed, and both the binding affinity and the binding kinetics to be accurately estimated. This method has been applied to simulate the binding events of the anti-Alzheimer’s disease drug (-)-Huperzine A (HupA) and E2020to their target Torpedo california acetylcholinesterase (TcAChE), respectively. The computational results were all well agreed with our concurrent experimental measurements. Briefly, all of the predicted values of binding free energy and activation free energies of association and dissociation deviated from the experimental data only by less than1kcal/mol to HupA-TcAChE, and2kcal/mol to E2020-TcAChE. The results prove that our computational method for predicting drug-target binding kinetics is reliable and effective.In conclusion, this dissertation has developed several drug design methods based on drug binding thermodynamics, including a quantitative statistical thermodynamic model for describing the concentration-response relationship between activators and potassium ion channels, target-specific docking programs iDNASBinder and iMpSDock, and a flexible docking program iFitDock that considering both ligand and protein flexibilities and the interactions of explicit water molecules. These methods have solved some problems in current drug design methods, and enriched the technologies for the paradigm of the binding affinity based drug development. More importantly, an efficient and reliable computational method was developed for simulating drug-target binding kinetics, which provides a new tool for the binding kinetics based drug development.