QSAR And Molecular Design Studies Of Proteasome Inhibitors

One of the most important degradation pathways of most intracellular proteins isthe ubiquitin-proteasome system pathway. This pathway is closely related to a variety ofdiseases including tumor, multiple myeloma, neurodegenerative diseases, inflammation,myocardial infarction. So, design and synthesis of the proteasome inhibitors aimed atproteasome have been a hotspot in recent years. At present, the proteasome inhibitorBortezomib （Velcade） has been approved by FDA in2003as the treatment of multiplemyeloma drug. The second-generation proteasome inhibitors Carfilzomib have enteredthe FDA review stage. Up to date, there are many types of molecules which are foundwith proteasome inhibitory activities, i.e. aldehyde peptides, boric acid peptides,epoxyketones, beta-lactone, polyphenols, flavonoids and so on. Besides, there also somemolecules from plant, metal complexes and other new types of proteasome inhibitors.Although some inhibitors have been demonstrated with strong in vitro inhibitoryactivity, poor specificity and low bioavailability are the bottleneck in development ofthese inhibitors.Due to the complexity, flexibility, covalent bonding characters of the mostproteasome inhibitors, a lot of troubles have been brought to quantitativestructure-activity relationship （QSAR） and molecular designs studies of proteasomeinhibitors. In this thesis, the comparative molecular field analysis （CoMFA）,comparative molecular similarity index analysis （CoMSIA）, Topomer CoMFA,Hologram QSAR and molecular docking were employed to the studies on5proteasomeinhibitors dataset with different structural types, i.e. the tyropeptin boric acid peptides,epoxyketone peptides, vinyl sulfone peptides, aldehyde tripeptides and non-covalentdipeptide derivatives. Some new inhibitors with high_predicted inhibitory activities havebeen designed based on QSAR models with strong_predictive powers. The main resultsare listed as follows:①Tyropeptin boric acid tripeptides inhibitors: The main interactions between theproteasome and these inhibitors are hydrogen bonds and hydrophobic interactions. Theoptimal CoMSIA model obtained is comstucted by hydrogen-bond donor field,hydrophobic field, hydrogen-bond acceptor field and steric field. The number ofprincipal component, the coefficient of determination （R²）, standard deviation（S）, thedetermination coefficient of cross-validation （Q²） and R²_predare3,0.882,0.188,0.494, and0.660respectively. Based on this optimal CoMSIA model,5moleculars with high_predictive activities have been designed, of which the activities reach nM level.②Epoxyketone peptide derivatives: The optimal model obtained is a classic2-DQSAR model established by SVM method. The number of descriptors, the coefficient ofdetermination （R²）, standard deviation（S）, the determination coefficient ofcross-validation （Q²） and R²_predof the optimal model are13、0.681、0.572and0.641.respectively. The main factors affecting the activities of56epoxyketone peptidemolecules are electronic, topology, hydrophobicity and steric properties. Based on thismodel,18molecules with high_predicted activities have been designed. The activities of18molecules designed reach nM level, which are higher than that of the templatemolecule.③Vinyl sulfone proteasome inhibitors: The optimal model of is a classic2-DQSAR method established by SVM method. The number of descriptors, the coefficientof determination （R²）, standard deviation（S）, the determination coefficient ofcross-validation （Q²） and R²_predof this optimal model are11,0.581,0.516, and0.501respectively. The results show that the electronic, composition, and steric properties arethe main factors affecting the activities of56vinyl sulfone molecules. The results alsoshow a water bridge between beta6and beta5active sites of proteasome and the ligandcan not be ignored.④Aldehyde tripeptide proteasome inhibitors: Two optimal QSAR models areestablished:1) a Topomer CoMFA model: the number of principal components, thecoefficient of determination （R²）, standard deviation （S）, the determination coefficient ofcross-validation （Q²） and R²_predare3,0.856,0.500,0.676, and0.5777respectively;2) aHQSAR model, the number of principal components, the coefficient of determination（R²）, standard deviation（S）, the determination coefficient of cross-validation （Q²） andR²_predare6,0.877,0.476,0.555, and0.8814respectively.⑤Non-covalent dipeptide proteasome inhibitors: The optimal model comes fromHQSAR modeling. The optimal debris combination of this model is A+B+C+H+Ch.The optimal fragments length is4-7and best hologram length is59. The number ofprincipal components, the coefficient of determination （R²）, standard error （SEE）, thedetermination coefficient of cross-validation （Q²） and R²_predare3,0.886,0.361,0.565,and0.6815respectively. Results show that the non-covalent interactions affecting theinhibory activity between the proteasome β5active site and a dipeptide proteasomeinhibitor are hydrophobic, hydrogen-bond and electronic interaction.