Prediction of Activities of Halogenated 2,4-Diphenyl Indeno [1,2b]pyridinol Derivatives Using QSAR Model Against Breast Cancer

Kaixuan Wang, Hongzong Si

Abstract: Breast cancer is one of the most common cancers among women. It is the first killer of women, and the existing drugs are not enough to treat the disease. In this study, a quantitative structure-activity relationship model used for predicting the IC50 value of those compounds was built by the chemical structures of a class of halogenated 2,4-diphenyl indeno[1,2b] pyridinol derivatives. All the 28 compounds were randomly split into a training set with 21 compounds as well as a test set with 7 compounds. The heuristic method in CODESSA program was utilized to optimize 28 compounds and establish linear models. Furthermore, four descriptors were selected and used to build a nonlinear model by gene expression programming method. The correlation coefficients R2, R2cv, S2 in the heuristic method were 0.671, 0.520 and 0.079, while in gene expression programming, the R2 and S2 were 0.728, 0.055 in the training set and 0.688, 0.062 in the test set, respectively. Both the two methods had good prediction performance. In comparison, the gene expression programming method was more consistent with the experimental values. The nonlinear model was supposed to the design and development of targeted anti-breast cancer drugs.

: Download (free)  CSE-PDF

QSAR Models of 2, 4-Disubstituted Imidazopyridines as Hemozoin Formation Inhibitors

Jieru Wang, Yuliang Li

Abstract: A new type of hemozoin formation inhibitors are the 2,4-disubstituted imidazopyridines, which can be used to design and develop anti-malarial drugs. Establishing a quantitative structure-activity relationship (QSAR) model can predict the physical and chemical properties of molecules. We used CODESSA software to find suitable molecule descriptors by heuristic method. Then a linear QSAR model with correlation coefficient (R2), square of standard error (S2) and the square of cross-validation coefficient (R2cv) are 0.7, 0.54 and 0.25 respectively. In addition, we randomly divided the 60 compounds into 45 training sets and 15 test sets to establish the nonlinear model using gene expression programming (GEP), the R2 and mean square error (MSE) of the training set are 0.92 and 0.17, and the test set are 0.85 and 0.10. It can be seen that the nonlinear results are significantly better than that of the linear results. It is hoped that this model will help in the design of hemozoin formation inhibitors

: Download (free)  CSE-PDF

A Stochastic Trust Region Method for Stochastic Optimization Problem with Equality Constraints

Youmeng He, Dan Xue, Donglei Liu

Abstract: A stochastic trust region algorithm is proposed to solve constrained minimization problems with stochastic objectives. This method can be used to deal with nonconvex problems. The new iterate is generated by minimizing an exact penalty function, whose size is controlled by a trust-region type parameter. The stochastic gradient is used to take the place of deterministic gradient for the determination of descent directions of the penalty function. The convergence of the method is discussed under some reasonable conditions. Some preliminary numerical results show that our method can effectively solve the stochastic nonlinear programming problems.

: Download (free)  CSE-PDF