New Chemometric Algorithms In Quantitative Structure-activity Relationship Studies

The research work in this thesis focuses on the QSAR studies and the design of some new chemometric algorithms used in this field.A version of modified particle swarm optimization (PSO) algorithm has been proposed. The PSO algorithm has been modified to adopt to the discrete combinatorial optimization problem and reduce the probability of sinking into local optima. The modified discrete PSO algorithm is proposed to select variables in MLR and PLS modeling and to predict antagonism of angiotensin II antagonists. Experimental results have demonstrated that the modified PSO is a useful tool for variable selection which converges quickly towards the optimal position.Many methods for training ANN tend to converge to local optima and may also lead to overfitting. In this, thesis, a hybridized particle swarm optimization approach was applied to neural network structure training (HPSONN). The continuous version of PSO was used for weight training of ANN, and the modified discrete PSO was applied to find appropriate network architecture. Network structure and connectivity are trained simultaneously. The two versions of PSO can jointly search the global optimal ANN architecture and weights. A new objective function is formulated to determine the appropriate network architecture and optimum value of the weights. The proposed HPSONN algorithm was used to predict carcinogenic potency of aromatic amines and biological activity of a series of distamycin and distamycin-like derivatives. The results were compared to those obtained by PSO and GA training in which network architecture was kept fixed. The comparison demonstrated that the HPSONN is a useful tool for training ANN which converges quickly towards the optimal position and can avoid overfitting in some extent.Piecewise modeling by particle swarm optimization (PMPSO) approach is applied to QSAR study. The minimum spanning tree is used for clustering all compounds in training set: to form a tree and the modified discrete PSO is applied to divide the tree for finding satisfactory piecewise linear models. A new objective function is formulated for searching the appropriate piecewise linear models. The proposed PMPSO algorithm was used to predict antagonism of angiotensin II. The results demonstrated that the PMPSO is useful for improvement of the performance of regression models.A new version of ant colony optimization (ACO) algorithm has been proposed. The modified ACO algorithm is proposed to select variables in QSAR studies and to predictinhibiting action of some diarylimidazole derivatives on cylcooxygenase (COX) enzyme. As a comparison to this method, the evolution algorithm (EA) was also tested. Experimental results have demonstrated that the modified ACO is a useful tool for variable selection that needs few parameters to adjust and converges quickly towards the optimal position.The variable selection in QSAR studies by MLR and PLS modeling has been performed using the evolution algorithm (EA). The Cp statistic has been modified and used as the objective function in the EA search for different combinations of molecular descriptors. The proposed procedures were used for the prediction of carcinogenicity of aromatic amines. QSAR study of 1 -phenylbenzimidazoles as inhibitors of platelet-derived growth factor receptor (PDGFR) has been performed. Some new electronic parameters Qo, Qm and Qp were suggested for characterizing effect of substituents. Descriptor Qm is shown to be an important variable to express effect of substituents.A version of orthogonal signal correction (OSC), ridge OSC (ROSC) algorithm has been proposed to improve the performance of the OSC. The problem of possible removal of useful information in OSC can be overcome in a more or less extent by using a ridge parameter in construction of the estimator in ROSC. Generalized cross-validation method was employed to select the ridge parameter and the number of OSC components. The proposed methodology has been tested in PLS modeling for QSAR studies of cyclooxygenase-2 inhibitors. It has been demonstrated that dat...