QSPR Study of the Complex Formation Constants between β-cyclodextrin and Some Organic Compounds

Cyclodextrins (CDs) are a group of structurally related natural products and also known as cycloamylosis . Recently CDs have been utilized in many different fields such as catalysis,separation science and technology, drug delivery, pharmaceutical application, food, personal careproducts and etc[1]. The purpose of this study is to construct a quantitative structure-propertyrelationship (QSPR) model that is able to predict the stability between different guest moleculesand β-cyclodextrin. This study is performed using the bee algorithm (BA) and the adaptiveneuro-fuzzy inference system (ANFIS). The 3-D structures of 230 compounds [1] were optimizedusing HyperChem software (version 8.0) with semi empirical AM1 optimization method. Afteroptimization a total of 3224 0-, 1-, 2-, and 3-D descriptors were generated using Dragon software(version 3.0) [2]. In the first, bee algorithm program was written in Matlab in our laboratory bythe authors and then was used to select the most important descriptors. Descriptor selectionprocedure starts with flying of n scout bees toward N-dimensional search space of N descriptors[3].Then the formation constants and error values are calculated using selected descriptors andmultiple linear regression model. Finally, the best descriptors are selected due to the lesscalculated errors. Therefore on the basis of BA, five descriptors were selected and applied asinput to the network of the ANFIS. Finally, to evaluate the predictive power of bee-ANFIS theoptimized model was applied to all dataset (training, test and validation sets). RMSEs of 0.2995,0.4213 and 0.3644 were obtained for the training, test and validation sets, respectively. Thecorrelation of coefficient were obtained as 0.9427, 0.8710 and 0.9275 for training, test andvalidation sets, respectively.

Keywords:

β-Cyclodextrin , QSPR , Bee Algorithm , ANFIS