Font Size: a A A

Establishment Of GPR40 Agonist Prediction Model Based On Deep Learning And Its Application On Virtual Screening

Posted on:2024-04-30Degree:MasterType:Thesis
Country:ChinaCandidate:J M YangFull Text:PDF
GTID:2544306917990079Subject:Medicinal chemistry
Abstract/Summary:PDF Full Text Request
[Objective]The activation of G protein-coupled receptor 40 can promote glucosedependent insulin secretion,making it an important target for the treatment of type 2 diabetes.Additionally,G protein-coupled receptor 40 is superior to other hypoglycemic drugs in preventing hypoglycemia.Although various drugs have entered clinical trials,they have not been approved yet,making GPR40 an important direction for drug development.However,relying solely on experimental screening for new active compounds is time-consuming and costly.Combining virtual screening methods can accelerate the discovery and development of active compounds.In this study,we first established a high-performance ensemble model,and then conducted virtual screening based on TCMTW library and the Chemdiv library using ensemble model,molecular docking,MM/GBSA,clustering analysis,and molecular dynamics simulation.Finally,we analyzed the substructural characteristics of GPR40 agonists based on PySmash.These studies provide a research foundation for the discovery and development of GPR40 agonists and the development of therapeutic drugs for type 2 diabetes.[Methods]In this study,we first established a new dataset and computed molecular representations,including 12 molecular fingerprints and 13 molecular descriptors.Next,we constructed an ensemble model and optimized its performance.Subsequently,we selected the Chemdiv library and TCMTW library as screening target libraries and conducted virtual screening based on the ensemble model,molecular docking,MM/GBSA,clustering analysis,and molecular dynamics simulation.Finally,we analyzed and discussed the substructural characteristics of G-protein coupled receptor 40(GPR40)agonists targeting the intrahelical allosteric site using PySmash.[Results]1.A high-performance ensmeble model was established,with a ROC AUC of 0.9496 on an external dataset.The code and data is available on the Github website(https://github.com/Jiamin-Yang/ensemble_model).2.Eight compounds were identified through virtual screening using the ensemble model,molecular docking,MM/GBSA,clustering analysis and molecular dynamics simulations,and they were found to stably bind to G protein-coupled receptor 40 in molecular dynamic simulation.3.Using PySmash,five characteristic substructures were identified,including a carboxyl group,a tricyclic substructure,a triple-bond substituted group,two conjugated systems connected by two atoms,and a trimethylbenzene-like structure.These substructures play specific roles in the conformational binding of ligands to the protein.[Conclusion]We established a high-performance ensemble model,which provides suggestions for the establishment and optimization of other ensemble models.We performed virtual screening using the ensemble model,molecular docking,MM/GBSA,clustering analysis and molecular dynamics simulations,and discovered eight potential G protein-coupled receptor 40 agonists.Our screening process for G-protein-coupled receptor 40 agonists also serves as a valuable reference for other virtual screening methods...
Keywords/Search Tags:GPR40, ensemble model, virtual screen, molecular docking, molecular dynamics
PDF Full Text Request
Related items