Computer Modeling On Inhibitors Of Acyl-Coenzyme A: Cholesterol Acyltransferase

Atherosclerosis is a major cause of cardiovascular disease, which is the mostsignificant causes of death of the old people in recent years. Treatment with statins is thewell-established frst choice for restraining atherosclerosis and hypercholesterolemia,howerver the complications and adverse reactions will appear. To further risk reduction, thedesired for new pharmacological agents that target specific steps of atherosclerosis hasintensified significantly. Acyl-coenzyme A: cholesterol acyltransferase (ACAT, E.C.2.3.1.26) is an allosteric enzyme that catalyzes the acylation of cholesterol to cholesterylesters with free cholesterol and long chain fatty acids. This is a promising target for theprevention and treatment of atherosclerosis diseases and hypercholesterolemia. In mammalsthe presence of two ACAT isoforms with distinct functions. A large number of experimentsimplies that ACAT-2may be sufficient as a potential target for the treatment ofatherosclerosis atherosclerosis and hypercholesterolemia. In this thesis, we built usingstructure–activity relationship (QSAR) models to study the selectivity of the ACATsinhibitors and predict the bioactivity of ACAT-2inhibitors.In the first part of this thesis,10molecular descriptors were selected to build ACATclassification models. Based on these descriptors, four models had been built throughdifferent classification methods of Kohonen’s Self-Organizing Map (SOM) and SupportVector Machine (SVM). As it turned out, the predictive accuracies of four models werehigher than85%for the test set, which reflected good robustness of two methods. The SOMand SVM models obtained in this study could be used for further virtual screening researchof selective inhibitors of ACAT for the development of a new anti-atherosclerotic agent. Inaddition, we summarized the structure-activity relationship of selective ACAT inhibitors,several substructures were found to play important roles in selectivity of inhibitors toACATs. It was helpful to find out the difference about the selectivity of inhibitors ofACAT1and ACAT2.In the second part of this thesis, two quantitative models for predicting the bioactivityof ACAT2inhibitors were built by (MLR) and (SVM) with a data set of compounds (95inhibitors). A abroad range of descriptors (469ADRIANA.Code descriptors) had beeninvestigated. And11selected descriptors can reflect the fundamental structure andcharacteristics of a molecule in detail, and it turned out that these chemical properties of amolecule (charge, electronegativity) were all important in modeling the bioactivity of ACAT2inhibitors. And according to our prediction results, these two methods performedequally well. These models will be useful in compound discovery and design of novelACAT2inhibitors.In summary, the above two aspects of the study were obtained good results. It is helpfulto find new ACAT2inhibitors using the classification models and prediction models webuilt.