Design And Verification Of Immune Checkpoint Inhibitors Based On Computational Model

In the tumor microenvironment,tumor cells can bind its programmed cell death protein ligand-1（PD-L1）to programmed cell death protein-1（PD-1）on the surface of T cells,avoiding the immune-attack of T cells and resulting in immune escape.Therefore,blocking the interaction between PD-1 and PD-L1 can be an effective means to treat cancer.However,it has been reported that there are fewer small molecule inhibitors targeting the PD-1/PD-L1 pathway.In order to develop PD-L1inhibitors efficiently,this study established a small molecule inhibitor automatic design system based on genetic algorithm and SVM model,and two new candidate compounds were selected as PD-L1 candidate inhibitors based on the evaluation results of molecular docking and molecular dynamics simulation.Furthermore,in vitro cellular and in vivo biological verification experiments were carried out,which confirmed that candidate compounds could block the PD-1/PD-L1 binding and exerted the anti-tumor activity.This dissertation is divided into the following six parts:1.OverviewThis chapter mainly focused on the research progress of immune checkpoint PD-1/PD-L1 pathway and computer models in terms of drug development.First,the PD-1/PD-L1 signalling pathway and the research progress of related small inhibitors were summarizes;then machine learning algorithms such as support vector machines and genetic algorithms,as well as molecular docking and molecular dynamics simulations were reviewed.Thus,the application in drug virtual design and screening could provide technical support for follow-up research.Finally,the scientific problems and research ideas to be solved in this dissertation were summarized.2.Design of PD-L1 small molecule inhibitors based on machine learning modelsThis chapter aimed to construct an automatic design system combining genetic algorithms and SVM regression models,which could design the novel PD-L1 small molecule inhibitors.Based on 1385 known small molecule inhibitors that target PD-L1,an SVM regression model for predicting the activity of PD-L1 small molecule inhibitors was successfully constructed.This model was optimized through the grid search algorithm,and the optimal parameters of SVM activity prediction model were determined to be the cross-validation mode as 7-Fold,showing that the kernel function type was the RBF kernel function,the penalty factor C was 10 and the parameter gamma was 0.01.After the optimization,the model performance was:RMSE=0.634,R2=0.806.Using the constructed PD-L1 inhibitor activity prediction SVM regression model as a fitness assessment method,a PD-L1 small molecule inhibitor automatic design system was successfully established.Setting the initial population size of the genetic algorithm as 100,the gene mutation probability as 0.01,and the genetic algebra as 1000,and finally 470 compounds with better predicted activities were automatically designed.3.Evaluation of PD-L1 small molecule inhibitors based on molecular simulationThis chapter aimed to screen new candidate compounds through molecular docking and molecular dynamics simulation.First,the interaction mode of the positive control compound BMS202 and the PD-L1 dimer complex（PDB ID:5J89）was established by kinetic simulation technology,followed by the binding evaluation of the PD-L1 dimer protein and the candidate compounds.The key residues in the active center were identified:THR20,ILE54,VAL55,TYR56,TRP57,MET115,ILE116,SER117,TYR118,ALA121,Asp122,TYR123,Lys124 and Arg125;also,other interaction patterns were revealed,including hydrogen bonds,π-πstack Interaction modes,and salt bridge.All these interaction patterns were applied as a reference docking mode for the binding of protein and candidate molecules.Secondly,26 new compounds with affinity<-12kcal/mol were screened through the flexible docking of Autodock vina,and among these candidates,two compounds,PD-1/PDL1-Ser and PD-1/PDL1-Ser-OEt,were selected.Finally,the binding patterns of each candidate,either PD-L1-Ser or PD-L1-Ser-OEt,and the PD-L1 dimer complex were studied separately through molecular dynamics simulation technology,and in view of MM/PBSA binding free energy analysis,candidate compounds could effectively inhibit the interaction of PD-L1/PD-1 interactions due to these candidates stably bound to PD-L1 dimer.Therefore,two candidate compounds were verified as the potential PD-L1 inhibitors.4.Synthesis of PD-L1 small molecule inhibitorsThis chapter aimed to synthesize candidate compounds PD-L1-Ser and PD-L1-Ser-OEt for the subsequent evaluation of drug efficacy.First of all,the reverse synthesis analysis of PD-L1-Ser and PD-L1-Ser-OEt was conducted to determine their synthetic schemes.Then,these compounds were synthesized and identified by hydrogen nuclear magnetic spectrum,carbon spectrum and mass spectrometry.5.In vitro activity evaluation of PD-L1 small molecule inhibitorsThis chapter aimed to investigate the in vitro biological activities of candidate compounds PD-L1-Ser and PD-L1-Ser-OEt.First,the homogeneous time-resolved fluorescence resonance energy transfer（HTRF）technique was used to investigate the inhibitory activity of two candidate compounds on PD-L1 binding.The results showed that compounds PD-L1-Ser and PD-L1-The inhibitory activity of Ser-OEt was 50.02%and 67.41%at a concentration of 1μM,and the related inhibitory activities were similar to that of the positive control BMS202 at the same concentrations,indicating that both candidate compounds had the ability to block PD-1/PD-L1 binding in vitro.Secondly,the anti-tumor activity of the compounds PD-L1-Ser and PD-L1-Ser-OEt were investigated using human breast cancer cell MDA-MB-231 and mouse breast cancer cell 4T1.The results showed that PD-L1-Ser-OEt had a strong anti-tumor effect on MDA-MB-231 and 4T1 cells,showing IC₅₀values of 338.5μg/m L and 174.4μg/m L,respectively.However,the cellular cytotoxicity of PD-L1-Ser was weak,showing the IC₅₀values were 402.2μg/m L for MDA-MB-231 cells and 303.6μg/m L for 4T1 cells,respectively.6.Study on the in vivo anti-tumor activity of PD-L1 small molecule inhibitorsIn this chapter,candidate compounds PD-L1-Ser and PD-L1-Ser-OEt were investigated in a 4T1-tumor-bearing mouse model in terms of their immune and anti-tumor effects in vivo.The results showed that the anti-tumor effect of the PDL1-Ser-OEt administration group was significantly better than that of the positive compound 5-FU,and its tumor inhibitory rate was up to 93.81%.The results of IFN-γand IL-4 ELISA evaluation showed that the serum IFN-γand IL-4 concentrations were increased in the PD-L1-Ser and PD-L1-Ser-OEt administration groups,indicating that candidate compound had an in vivo immune checkpoint activity.In summary,this study built a computational model based on machine learning and molecular simulation,which could efficiently design novel PD-L1 small molecule inhibitors,and could provide a research tool and were potential for the development of small molecule immune checkpoint inhibitors.