DOX:A New Computational Protocol For Accurate Prediction Of The Protein-Ligand Binding Structures

To gain a deep understanding of the properties,functions,and mechanisms of the protein-ligand complexes,the detailed knowledge of the protein-ligand binding structures is always required.Accurate protein-ligand binding structures is also a prerequisite for structure-based drug design.This dissertation focuses on accurate theoretically prediction of the protein-ligand binding structures and mainly includes three parts as follows.1.Constructing a new computational protocol for accurate prediction of the protein-ligand binding structures:For the complex protein-ligand molecular systems,accurate theoretically prediction of the right binding structures with lower energy states depends on the ability to realize the global searching at the complex potential energy surfac and local optimization at quantum mechanism level of complex molecular conformations with lower energy states.In the current study,we have put forward a novel strategy of searching binding conformations with lower energy states across multiple-levels potential energy surfaces(BC-SAMP)to accurately predict of the protein-ligand binding'structures.Then we have put forward a new DOX protocol to realize the attempt of BC-SAMP strategy.2.To explore the prediction ability of DOX protocol in biological system,we have carried out two parts of work as follows:(1)We set up a test set consisting of 15 crystal structures for the human 3-hydroxy-3-methylglutaryl coenzyme A reductase(HMGR)to test the accuracy of our DOX hybrid computational method.The primary problem that the DOX protocol needs to address is the quality of the sampling space of the protein-ligand binding conformations(SSBC).We use molecular docking to generate the SSBC and examine the performance of DOX protocol based on native-docking,re-docking and cross-docking.From native-docking to cross-docking,the difficulty of the study gradually increased at the same time gradually approaching the real situation.Based on native-docking,with the binding geometries of individual ligands in the crystal structure and the energy of the crystal complex as the standard,DOX protocol can efficiently predict accurate protein-ligand binding geometries for the test set with an average RMSD of 0.54 A and an average RBE of 1.13 kcal/mol.At the re-docking level,we use Confort to generate the conformation of ligand for molecular docking.Because of the structural complexity of the ligands in the test set,we find that when the conformation of ligand is not derived from the crystal structure but from a low energy conformation produced by Confort,will have a great influence on the quality of SSBC generated by molecular docking.We have used Confort to generate six low-energy conformations and increasing the number of binding conformations in SSBC to solve the problem.The number of protein-ligand binding conformations increases will result in an increase in the computational complexity of the ONIOM3 calculation.We have used PM7 calculation to replace the ONIOM3 calculation to balance the computational accuracy and efficiency.We name the DOX protocol Docking-ONIOM3-X02 as DOX1.0,the Docking-PM7@MOPAC-X02 as DOX2.0.In re-docking the DOX2.0 protocol can also efficiently predict accurate protein-ligand binding geometries for the test set with an average RMSD of 0.50 A and an average RBE of 0.92 kcal/mol.Although the re-docking is more strict than native-docking,the performance of DOX2.0 protocol based on re-docking is slightly better than DOX1.0 protocol in native-docking,indicating the benefit of the updated DOX protocol.For the validation of the universality of DOX protocol,we also test the DOX2.0 protocol against another 13 protein-ligand complexes involving three proteins,including scytalone dehydratase(SDase),Neuraminidase(NA)and 1,3,8-trihydroxynaphthalene reductase(3HNR).The obtained results show that for all 13 protein-ligand complexes without an exception the DOX2.0 protocol meet the prediction accuracy with an average RMSD less than 1.0 A and an average RBE less than 2.0 kcal/mol.At the cross-docking level,DOX2.0 protocol can also efficiently predict accurate protein-ligand binding geometries for the 15 human HMGR-statin complexes with an average RMSD of 0.63 A,indicating that in practical application DOX2.0 protocol can accurately predict the protein-ligand binding geometries for this system.(2)We use light-harvesting protein C-phycocyanin from cyanobacteria to examine the performance of DOX2.0 protocol on covalently bound protein-ligand complexes.It is reported that phycocyanobilin adopts a cyclic conformation in the free state and a stretch conformation in protein-ligand complex.We need the conformation of the phycocyanobilin in its free state as the starting conformation of the molecular docking for DOX2.0 protocol.So we first search the conformation of phycocyanobilin based on SAMP strategy.The results show that SAMP strategy can search for the cyclic conformation of phycocyanobilin molecule,indicating that SAMP strategy could be applied potentially to small molecule conformation search.Then,the binding structure of phycocyanin and phycocyanobilin is predicted by DOX2.0 protocol.With the binding geometry of ligand in the crystal structure and the energy of the crystal complex as the standard,DOX protocol can efficiently predict accurate phycocyanin-phycocyanobilin binding geometry with a RMSD of 0.31 A and a RBE of 1.97 kcal/mol,preliminary indicating that DOX protocol could be applied potentially to covalently bound protein-ligand complexes.3.Application of DOX protocol in the design of human liver fructose-1,6-bisphosphatase inhibitors:First we carry out a virtual screening study for the AMP site of Hu-FPBase.After virtual screening from Specs database we obtain 25 compounds,and one of them showed a micromolar inhibitory activity.As we have not obtained the crystal structure of the lead compound and Hu-FBPase at present,so we can not get the binding structure information of the lead compound and Hu-FBPase by the experimental method.If we base on the lead compound itself to optimize the structure,it has a certain blindness.As the DOX protocol has been successfully tested in HMGR,3HNR,SDase and NA systems,indicating that it has a certain degree of reliability in accurately predicting the binding structures of protein-ligand.So we use DOX2.0 protocol to predict the binding structure of Hu-FBPase and the lead compound,then according to the predicted binding structure to optimize the lead compound in order to get compounds with higher inhibitory activity against Hu-FBPase.This study suggests that the DOX protocol might be a promising computational strategy and route to lead optimization in structure-based drug design.