Design And Molecular Simulations Of Anti-cancer Drugs For Kinases

Malignant tumors are non-infectious fatal diseases,and are recognized as one of the two major killers of human health with cardiovascular diseases.Molecular targeted tumor therapy has become one of the mainstream research directions in the field of malignant tumor therapy due to its high efficacy,low toxicity and high specificity.Kinases are the important targets in the research and development of anti-cancer drugs,such as ALK/ROS1,a key target for non-small cell lung cancer,JAK2,an important target for many malignant hematological malignancies,and CDK 9,a popular target for hematological malignancies and lymphoma.Currently,important breakthrough has been made in the development of new drugs targeting ALK/ROS1,JAK2,and CDK9.A number of small molecule inhibitors of these kinases have been used for clinical treatment or clinical research.However,with the widespread use of kinase inhibitors,clinical application problems such as acquired drug resistance and toxic side effects have emerged.In this thesis,a variety of molecular simulation methods have been used to conduct in-depth theoretical studies on the binding,acquired resistance and selectivity mechanism of the small molecule inhibitors targeting these kinases,which is anticipated to provide valuable guidance to screen and design novel and effective kinase inhibitors.In Chapters Two and Three,we employed a combined strategy,including molecular docking,molecular dynamics(MD)simulations,MM/GBSA binding free energy calculations and decomposition techniques,to reveal the binding mechanism of piperidine carboxamides Type-I1/2 ALK inhibitors.Based on the simulation results,a series of 1-purine-3-piperidinecarboxamide Type-I1/2 ALK inhibitors were designed,and five compounds showed the IC50 values of picomolar level.In particular,001-017 showed strong inhibitory activity against crizotinib and ceritinib-resistant cell lines with the L1196 M,C1156Y,R1275 Q,or F1174 L ALK mutant.Subsequently,ensemble docking and MD based MM/PB(GB)SA calculations were used to mimic the “conformational selection” and “induced-fit” phenomena during the binding process of these Type-I 1/2 inhibitors,respectively.The simulation results showed that the incorporation of protein flexibility was critical to accurately predict the binding affinity of the 1-purine-3-piperidinecarboxamide Type-I1/2 inhibitors.Besides,umbrella sampling(US)simulations showed that 001-017 had the strongest binding affinity and the longest residence time.In Chapter Four,MD simulation,US simulation,MM/GBSA binding energy calculation and decomposition techniques were used to illuminate the resistance mechanism of the L884 P mutation resistant to BBT594 and CHZ868.The US simulations revealed that the L884 P mutation enhanced the flexibility of the allosteric pocket,especially the ?3-strand,?C-helix and DFG motif,which was supported by the increased conformational entropy(-T?S)and RMSFs.Energy decomposition analyses suggested that L884 P mutation impaired the interaction between BBT594 and Tyr931 as well as Glu898.In Chapter Five,MD simulations,steered molecular dynamics(SMD)simulations,US simulations,MM/GBSA free energy calculations and decomposition methods were used to reveal the selectivity mechanism of 4-(thiazol-5-yl)-2-(phenylamino)pyrimidine-5-carbonitrile derivatives toward CDK9 over CDK2.The simulation results showed that the intrinsic flexibility of CDK9 and the “spacious” binding pockets facilitated the binding of the bulky compounds 12 u and 4.Energy decomposition indicated that the non-conserved residues surrounding the binding pocket were also important to determine the inhibitor selectivity.In Chapter Six,we constructed the virtual screening strategies based on the multiple protein structures of ROS1,and evaluated their screening accuracy for ROS1 inhibitors.Firstly,the na?ve bayesian classification(NBC)technique was used to integrate the docking scores for multiple structures.The results showed that the NBC model based on the docking results of ten ROS1 structures had the highest prediction accuracy.In addition,we combined the MM/GBSA residue decomposition energy term(MIEC)and support vector machine(SVM)to construct a MIEC-SVM prediction model for ROS1.And then the MIEC-SVM model showed superior inhibitor enrichment and binding affinity ranking capabilities than Glide-XP.