QSPR/QSAR Studies In Chemistry, Medicinal Chemistry And Environmental Science

Quantitative structure-property/activity relationships (QSPR/QSAR) methods, as one of the most important research fields, have been widely used for the prediction of various physicochemical properties and biological activities of organic compounds by using different statistical methods and various kinds of molecular descriptors. Due to the rapid development of computer science and its extensive application in chemistry, the studies of QSPR/QSAR met a new age. The main objects that QSPR/QSAR studied are biological activities, toxicity of organic compounds, and various physicochemical properties and have been widely used in chemistry, medicinal chemistry, environment chemistry and drug design.There is some relationship between molecular structures and properties/activities. We can deal with there liner or nonlinear relationships by constructing QSPR/QSAR models through various machine learning methods, such as heuristic method (HM), multiple linear regression (MLR), artificial neural network (ANN), support vector machine (SVM), least-squares support vector machine (LS-SVM), and projection pursuit regression (PPR).What QSPR/QSAR mainly studied are orgnic molecules or drug ligands in docking. If we want to study biological macromolecules (or drug acceptors), however, it is hard for QSPR/QSAR methods to do these studies. In addition, complicated phenomena in chemistry are controlled by a series of physical laws, which are part of quantum mechanics. The equations of quantum mechanics are very hard to solve for real systems. Fortunately, the advent of powerful computers made it possible to make predictions. However, this computational success leaves open the main question of finding a reliable link between chemical concepts and quantum mechanics. The former have been developed over time based on phenomena observable at the human level. The theory of Quantum Chemical Topology (QCT) represents a bridge between the concepts used to explain chemistry and the highly complicated wave function of quantum mechanics. QCT uses p(r) as a main source of information. Multipole moments in the representation of atomic electron density are vital for modeling electrostatic interactions accurately.In Chapter 1, a brief description of the QSPR/QSAR principle, research methods and status was presented. Much emphasis was put on the realization process of QSPR/QSAR. In this section, we also introduced the SVM, LS-SVM, and PPR and an overview of the application in QSPR/QSAR field. In addition, QCT theory and its application were also introduced in this chapter.In Chapter 2, we concluded the application of LS-SVM in the QSPR/QSAR studies. A brief description was given below:(1) Prediction of volatile components retention time in blackstrap molasses by least-squares support vector machine. Heuristic Method (HM) and LS-SVM were used to develop the linear and nonlinear QSPR models between the retention time (RT) and five molecular descriptors of 45 volatile compounds. The root mean-square errors (RMSE) in RT predictions for the test data set given by HM and LS-SVM were 2.193 and 2.728, respectively, which showed the performance of LS-SVM model was better than of the HM model. The prediction results are in agreement with the experimental values very well. This paper provided a new and valid way to predict the logRT values of the volatile compounds.(2) Classification study of novel piperazines as antagonists for the melanocortin-4 receptor based on least-squares support vector machines.5 descriptors calculated solely from the molecular structures of compounds by forward stepwise linear discriminant analysis (LDA) were used as inputs of the LS-SVM model. The accuracy of training set for LS-SVM was 97.62% and the test set was 95%. It gave a useful and alternative way for classification of the activity of MC4 selective inhibitors.In Chapter 3, we concluded the application of PPR in the QSPR/QSAR studies. A brief description was given below:(1) QSAR models for 79 CCR5 receptors binding affinity of substituted 1-(3,3-diphenylrpopyl)-piperidinyl amides and ureas based on HM, SVM and PPR. A subset of eight molecular descriptors selected by HM in CODESSA was used as inputs for SVM and PPR. The results obtained by nonlinear SVM and PPR models were compared with those obtained by the linear heuristic method. The prediction results of the SVM and PPR models were better than that obtained by HM. The models of SVM and PPR led to a correlation coefficient square (R2) of 0.732 and 0.726 and root mean-square error (RMSE) of 0.210 and 0.207 for the test set and the values for HM model were 0.715 and 0.238 respectively.(2) Prediction of photolysis of PCDD/Fs adsorbed to spruce (Picea abies (L.) Karst.) needle surfaces under sunlight irradiation based on projection pursuit regression. HM and PPR were used to develop linear and nonlinear models, respectively. Both the linear and the nonlinear model can give very satisfactory prediction results:the RMSE in prediction for overall data set were 0.042 and 0.032, and the R2 were 0.828 and 0.893, respectively. By the analysis of the proposed model, we can conclude that the factors influencing the photolysis half-life (t1/2) values of PCDD/Fs under sunlight irradiation mainly included the electrostatic features and stability of molecules. This paper proposed two rapid, effective methods to predict the t1/2 values of PCDD/Fs.In Chapter 4, concluded the application of QCT in the biological systems studies. A brief description was given below:Atom-atom electrostatic energy studies using high-rank topological multipole moments for protein crambin. We calculate the atom-atom elestrostatic energy using multipole moments which was based on the method of QCT. In crambin, there are 5 elements and in total 15 types of atom-atom interactions that were grouped based on the element. the minimum values of the energy multipole expansion in the convergent regions for all the fifteen types of atom-atom interactions are obtained. In addition, A comparison between AMBER results and QCT results was carried out to show that use of multipole moments can improve the accuracy of electrostatic energy at short-range. The next major task is to work out a force field which includes the multipole moments and polarization to get better energy results.