Application Of Computational Chemistry In Research Of Antitumor Drug And Prediction Of Atomization Energies

Quantitative structure-activity relationship (QSAR) methods, are the most promising and successful tools to provide rapid and useful meanings for predicting the biological activity or toxicities of organic compounds by use of different statistical methods and various kinds of molecular descriptors. The goal of QSAR is to develop models on a training set of compounds, these models will then allow for the prediction of the biological activity of related chemicals. This kind of study can not only develop a method for the prediction of the property of compounds that have not been synthesized but also can identify and describe important structural features of molecules that are relevant to variations in molecular properties, thus gain some insight into structural factors affecting molecular properties.In the current thesis, Chapter 1 included a brief description of the QSAR history, principle, realization process and research status. In Chapter 2, a systematic study of three dimensional quantitative structure activity relationship (3D-QSAR) on 99 Podophyllotoxin Derivatives was peformed with respect to their anticancer activity against KB cells in vitro through comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA), statistically reliable models were obtained with good predictive power. The predictive CoMFA and CoMSIA model for training set with cross-validated r~2(q~2)=0.644 and 0.764 , non-cross-validated r~2=0.930 and 0.964, standard error of estimate(s)=0.134 and 0.096. The predictive abilities of CoMFA and CoMSIA models were validated by a test set of 19 compounds with predictive r~2 (r stands for linear correlation coefficient) values of 0.706 and 0.683, respectively. In Chapter 3, a systematic study of 3D-QSAR on 38 Podophyllotoxin Derivatives which are designed and Synthesized by College of Chemistry and Chemical Engineering, Lanzhou University against A-549 cells in vitro through CoMSIA, statistically reliable models were obtained with good predictive power. The predictive CoMSIA model for training set with cross-validated r~2(q~2)=0.665, non-cross-validated r~2=0.993, the predictive abilities were validated by a test set with predictive r~2 values of 0.946. The obtained results of Chapter 2 and 3 can not only partly explain the QSAR of these podophyllotoxin derivatives but provide more beneficial guidance for designing new podophyllotoxin derivatives.Support vector machine (SVM) as a novel type of machine learning method, is gaining popularity and wide applications due to many attractive features and promising generalization performance. Based on the success of the literature of calculating molecular enthalpies using a combined method of B3LYP/6-311g (3df,2p) and SVM (B3LYP-SVM method), Chapter 4 of the dissertation is the calculation of atomization energies using the method, the overall mean absolute deviation of the B3LYP-SVM method from experiment for the 250 molecules is 2.079 kcal/mol and 12.70kcal/mol for the original B3LYP/6-311g (3df, 2p) calculations. This method of calculation on atomization energies for most familiar inorganic and small- to medium-sized organic compounds achieves the desired objectives of simple ,low cost ,high-precision and good generalization performance.