Screening Of Human Dihydroorotate Dehydrogenase Inhibitors

Human dihydroorotate dehydrogenase(hDHODH)is a therapeutic target associated with immunity and cancer,and inhibitors of this target have shown great potential for applications in the treatment of autoimmune diseases and cancer.Among them,inhibitors including leflunomide have been used as one of the clinical treatments for autoimmune diseases such as rheumatoid arthritis,systemic lupus erythematosus and malignant solid tumors.However,hDHODH inhibitors may produce some side effects,causing some diseases by interfering with DNA synthesis and inducing apoptosis.In addition,the long-term use of existing inhibitors can bring about drug resistance,which has limited the spread and application of current hDHODH inhibitors.Therefore,it is urgent to design a new kind of small molecule inhibitors.In the first part of this paper,the ASGBIE＿ESS method is used to calculate the binding free energy between hDHODH and AF derivatives.The correlation between the calculated and experimental values was as high as 0.907,indicating the superior performance of the method AGBIE＿ESS in the calculation of the binding free energy of hDHODH and its inhibitors.The alanine scan method is employed to calculate the contribution of individual residue to the binding free energy in hDHODH,identifying the key role of hot-spot residues in the ligand-protein binding process including M13,L16,Q17,H26,F32,F68,R106,Y326 and T330.This study combines molecular dynamics simulations with ASGBIE＿ESS and applies this method to virtual screening of inhibitors.Consequently,a series of new compounds with unique backbones are discovered.These compounds are able to form stable hydrophobic interactions with residues around the hydrophobic region pocket of the protein.In addition,it is discovered that the binding ability of those compounds is significantly improved when they form hydrogen bonding interactions with residues such as Q17 and R106 in the pocket.These compounds can serve as promising inhibitors of hDHODH.The process of virtual screening relies heavily on the databases but it is disadvantageous to conduct virtual screening based on commercial databases with patent-protected molecules,high molecular toxicity and side effects.Therefore,the second part of this paper utilizes generative recurrent neural networks(RNN)containing long short-term memory(LSTM)cells to learn the properties of drug molecules in the Drug Bank,aiming to obtain a new and virtual screening compound database with drug-like properties.Ultimately,a compound database consisting of 26316 molecules is obtained by this method.To evaluate the potential of this compound database,a series of tests are performed,including chemical space,ADME properties,molecular fragmentation,and synthesizability analysis.As a result,it is proved that the database is equipped with good drug-like properties and a relatively new backbone,its potential in virtual screening is further tested.Finally,a series of seedling compounds with completely new backbones are obtained through docking and binding free energy calculations.