Machine learning models to identify lead compound and substitution optimization to have derived energetics and conformational stability through docking and MD simulations for sphingosine kinase 1

Dhanabalan, Anantha KrishnanAnantha KrishnanDhanabalanJalgham, Ramzi T.T.Ramzi T.T.JalghamHaribabu, JebitiJebitiHaribabuGunasekaran, KrishnaswamyKrishnaswamyGunasekaran2025-10-102025-10-10202513811991; 1573501Xhttps://hdl.handle.net/20.500.12740/23360Sphingosine kinases (SphKs) are a group of important enzymes that circulate at low micromolar concentrations in mammals and have received considerable attention due to the roles they play in a broad array of biological processes including apoptosis, mutagenesis, lymphocyte migration, radio- and chemo-sensitization, and angiogenesis. In the present study, we constructed three classification models by four machine learning (ML) algorithms including naive bayes (NB), support vector machine (SVM), logistic regression, and random forest from 395 compounds. The generated ML models were validated by fivefold cross validation. Five different scaffold hit fragments resulted from SVM model-based virtual screening and docking results indicate that all the five fragments exhibit common hydrogen bond interaction a catalytic residue of SphK1. Further, molecular dynamics (MD) simulations and binding free energy calculation had been carried out with the identified five fragment leads and three cocrystal inhibitors. The best 15 fragments were selected. Molecular dynamics (MD) simulations showed that among these compounds, 7 compounds have favorable binding energy compared with cocrystal inhibitors. Hence, the study showed that the present lead fragments could act as potential inhibitors against therapeutic target of cancers and neurodegenerative disorders. © 2025 Elsevier B.V., All rights reserved.restrictedAccessCOMBINATORIAL CHEMISTRYFRAGMENT SCREENINGIFP ANALYSISMACHINE LEARNINGMOLECULAR DOCKINGMOLECULAR DYNAMICS SIMULATIONSSPHINGOSINE KINASESSPHKSSVMLEADPHOSPHOTRANSFERASESPHINGOSINE KINASEENZYME INHIBITORSPHOSPHOTRANSFERASES (ALCOHOL GROUP ACCEPTOR)R-CARET PACKAGEPF 543 INHIBITORPROTEIN INHIBITORSPHINGOSINE KINASE 1UNCLASSIFIED DRUGENZYME INHIBITORANGIOGENESISARTICLEBAYESIAN LEARNINGBINDING AFFINITYCRYSTAL STRUCTUREDEGENERATIVE DISEASEENERGY TRANSFERHUMANIC50LEARNINGMALIGNANT NEOPLASMMOLECULAR DYNAMICSMUTAGENESISPROTEIN STRUCTURERANDOM FORESTRECURSIVE FEATURE ELIMINATIONROOT MEAN SQUARED ERRORSUPPORT VECTOR MACHINECHEMISTRYHYDROGEN BONDMETABOLISMTHERMODYNAMICSHUMANSHYDROGEN BONDINGMOLECULAR DOCKING SIMULATIONMOLECULAR DYNAMICS SIMULATIONMachine learning models to identify lead compound and substitution optimization to have derived energetics and conformational stability through docking and MD simulations for sphingosine kinase 1Artículo https://doi.org/10.1007/s11030-024-10997-4