Computer Aided Study Of Regulators Based On Thyroid Hormone Receptors

Thyroid hormone(TH)is widely distributed in the human body,and it would regulate the growth,development and energy metabolism together with growth hormone.Studies indicate that the function of TH is mediated by binding to thyroid hormone receptors(TRs),which is a member of the NRs superfamily,and exhibiting crucial roles in development,homeostasis and many disease processes,in addition,the abnormal state of TRs is closely related to material metabolism and energy metabolism,such as obesity,hypercholesterolemia and diabetes,thus,emerging as major targets for pharmaceutical agonists and antagonists.Up to now,a number of diverse TRs ligands have been developed,however,these compounds exhibit side effects in the process of functional control.A large number of functional factors have been found in vegetables and fruits,therefore,in order to screen novel and potent TRs regulators,a series of computer aided calculation methods including classification method,two dimensional quantitative structure-activity relationship(2D-QSAR),three dimensional quantitative structure-activity relationship(3D-QSAR),molecular docking,molecular dynamics,inverse virtual screening method,and pharmacophore model is applied to study the TRs regulators,the aim is to solve the following issues:(1)Understand the binding mode between the regulators and TRs at molecular level,(2)Establish the corresponding model,and search the Natural Products Database to screening the ideal compounds that target TRs.The main findings can be described as follows in detail:(1)The statistical method C4.5,support vector machine(SVM)and random forest(RF)are employed on TRs agonists and antagonists to establish classification models,the results indicate that the C4.5 and RF models present more than 80% external prediction,and the predictive accuracy for the test set is more than 90%,indicating that the classification model can be used to predict the characteristics of the newly discovered thyroid hormone receptor regulators.(2)Multivariable linear regression(MLR),partial least squares regression(PLSR)and support vector regression(SVR)is used to develop the 2D-QSAR models based on regulators acting on the fist and the second TRs targets,these models are significant in statistics,and possess the capality to predict the activity of the test set,furthermore,the significant molecular descriptors affecting the TRs binding activity are identified.And the derived models can be employed to predict the activity of the novel TRs regulators.(3)Several computational models coupling with 3D-QSAR,molecular docking and molecular dynamics,are developed based on sulfonylnitrophenylthiazoles(SNPTs),indane analogs,thyroid hormone analogs,phosphonic acid derivatives and ?-aminoketone analogs,the results indicate that the derived models are satisfied in statistics,and the binding mode between regulators and TRs is clearly illustrated.(4)In the present work,an inverse virtual screening approach is performed to screen possible novel targets for dityrosine.Molecular docking studies are performed on a panel of targets extracted from the potential drug target database(PDTD).Finally,tubulin(-11.0 kcal/mol)is identified as a target for cis-dityrosine;several targets including tubulin(-11.2 kcal/mol),thyroid hormone receptor beta-1(-10.7 kcal/mol),and leukotriene A4 hydrolase(-10.2 kcal/mol)present high binding affinities for trans-dityrosine.The application of inverse virtual screening method may facilitate the prediction of unknown targets for known ligands,and direct future experimental assays.(5)The pharmacophore models are constructed on TR? and TR?,respectively,and the developed models are validated by molecular docking,results suggest that the two pharmacophore models are all stable.In addition,we apply the derived models to search for the potential compounds in the natural products database,and the ideal compounds are identified for TR?(bisdehydrostemoninine B and bisdehydrostemoninine A)and TR?(Artocarmitin A and Oxydihydroartocarpesin).In conclusion,these computational models possess high predictive performance,and can provide some insights into the structural characteristics that affect the ligand binding activity and provide some meaningful clues in the future experiment,and provide guidance for our daily diet.