Structure-Based Drug Design On Aurora Kinase And Predict Of PKa Values Of Aromatic Carboxylic Acids

It's financial and time-consuming to design a new drug.Computer models are available for drug design with less expense and time.It was observed that Aurora mitotic protein kinases(A,B and C) are frequently overexpressed in cells from various tumor types.Aurora-A kinase plays an very important role in the centrosome separation, maturation and bipolar spindle assembly during mitotic phase. It was found that Aurora-A is an important protein target for anti-cancer drug design.pKa(acid-base dissociation constant) is an important ADME properties of organic compunds,which is connected to drug's many other characteristics.The first part of this paper focuses on:(1) the similarity and the dissimilarity of Aurora-A and B;(2) drug design study based on the target Aurora-A.Firstly,the interaction of Aurora-A and its known inhibitors were computed by docking method.Some new scaffolds based on the known inhibitors were designed.Then a series of new compounds were designed,and they were screened by the properties of Lipinski's rule of 5. Afterwards,each molecule was docked into Aurora-A and the binding free energyΔG was calculated.From computation results,some new compounds with potential bioactivity were found.They might become new lead compounds and could be used for further experimental test.At last,the synthesis accessibility of these compounds was evaluated by the program SYLVIA.In the second part of this thesis,two Quantitative Structure-Activity Relationship(QSAR) models for the prediction of the pKa values of 180 aromatic carboxylic acids were developed.These molecules contain elements such as H,C,N,O,S,F,Cl,Br,and I with the molecular weight in the range of 122.12 to 288.34.The molecules were represented by 236 molecular descriptors calculated from the Cerius2 program.Twelve descriptors were selected with the statistical methods.The pKa values were predicted by a Multilinear Regression(MLR) analysis and a Support Vector Machine(SVM) Regression with 10-fold cross validation method. The model based on MLR analysis has a correlation coefficient of 0.90, and the standard deviation of 0.32 pKa units;the better model based on SVM Regression has a correlation coefficient of 0.91,and the standard deviation of 0.31 pKa units. In summary,in this thesis,the similarity and the dissimilarity of Aurora-A and B were investigated;based on the target protein Aurora-A, some new molecules with potential bioactivity were designed;two quantitative models for prediction of pKa values of aromatic carboxylic acids were built.