GSA Structural Protein-covalent Ligand-Binding Conformation Sampling Space

The protein ligand binding conformation is based on the key information required for structural drug design,which can be obtained through experimental and theoretical means such as crystal structure analysis,nuclear magnetic resonance,and molecular docking.However,experimental methods such as crystal structure often mean expensive costs and a very low success rate.Using modern molecular simulation technology to predict the binding mode of protein ligands is an important supplement to experimental methods.Among them,molecular docking is the most commonly used theoretical method,which can quickly predict the binding mode of drugs in protein binding sites.However,the existing molecular docking methods have limited accuracy in predicting the binding mode of protein ligands,and the error rate can reach about 50%,among which incomplete binding mode sampling accounts for a large proportion.In particular,it is worth mentioning that the molecular docking tools that can be used for covalent drugs are extremely limited and the results are not ideal.In recent years,covalent drugs have the characteristics of high efficiency and long action time due to the formation of covalent bonds with the target,and their remarkable therapeutic effects on certain diseases have attracted more and more attention.Covalent drugs occupy an increasingly important position in the drug market,and the demand for reliable computer-aided covalent drug design methods is increasing day by day,which is in sharp contrast with the deficiencies of existing methods.The theoretical prediction of protein ligand binding conformation mainly relies on the spatial matching and energy matching of protein ligands,which correspond to the conformation search method and scoring function in the molecular docking algorithm respectively.The error of molecular docking,especially the disappointing performance of covalent drugs,is quite a bit because the calculations did not search for the "correct conformation" at all,and it is impossible to talk about how to rank the "correct conformation" to the forefront.This paper proposes to use the Generalized Simulated Annealing Algorithm(GSA)to efficiently search the binding modes of covalent and non-covalent protein ligands,construct a rich and diverse binding conformation space,and conduct large-scale tests on different test sets to systematically investigate The feasibility and universality of the program.The main research work of this paper includes:First of all,for the sampling ability of the covalent binding mode of the GSA method,this article conducted a large-scale test on the COVDOX test set containing 405 protein-ligand complexes.The test results show that the sampling success rate of the GSA method can reach 86%(the probability that the total conformation sampling set contains at least one good conformation),and its performance basically does not change with the change of covalent bullets and residues,indicating that the GS A method is suitable for conformation search.It is feasible and has good universality.After combined testing with DOX,a protein ligand binding mode calculation method based on quantum chemistry,it was further found that the top ten conformations obtained by semi-empirical quantum chemistry PM7 optimization screening of the conformation set obtained by GSA have a high probability of at least one good conformation as high as 91%;After high-precision XO calculation,the TOP POSE prediction success rate is 80%.At present,the performance of mainstream covalent software fluctuates greatly on different test sets,and the success rate is less than 60%.Compared with the existing literature reports,the performance of the GSA method on various inhibitors is far superior to the existing covalent docking tools,which fully demonstrates that the method used in this article has significant advantages.Secondly,for the sampling ability of the non-covalent binding mode of the GSA method,this paper uses the CASF2016 test set to conduct a systematic test.The test set contains 57 different protein targets and a total of 285 inhibitor molecules.The test set not only has a variety of pharmacophores and rich inhibitor structures,but also has a huge amount of data,and the test results are reliable.The test results show that the sampling success rate of the GSA method can reach 88%.If it is combined with Surflex molecules to make up for each other,the total sampling success rate is even as high as 98%.In summary,the GSA method has good performance in constructing covalent and non-covalent combined sampling spaces.A complete sampling space is a crucial part of molecular docking,which lays a solid foundation for predicting accurate binding conformations.If combined with the high-precision binding energy evaluation method,it will be a very promising new method for predicting the binding conformation of protein ligands.