QSPR Study of Chemicals Adsorption on the MWCNT using Bee Algorithm as Descriptor Selection Method

Adsorption of chemicals on nanomaterials is an effective and important issue in nanotechnology. It can be studied as the main subject for complicated interactions between nanoparticles (NPs) with target molecules when they are used as drug delivery [1] diagnostic tools or environmental adsorbents. In this study, a new quantitative structure–property relationship (QSPR) methodology was used to model and predict the adsorption coefficients of some small organic compounds on multiwall carbon nanotube (MWCNT) based on bee algorithm (BA) method. The 3-D structures of these compounds were optimized using Hyper Chem software (version 7.0) with semi empirical AM1 optimization method. After optimization a total of 3224 0-, 1-, 2-, and 3-D descriptors were generated using Dragon software (version 3.0). A major problem of modeling is the high dimensionality of the descriptor space; therefore, descriptor selection is one of the most important steps. For descriptor selection in the first step, the bee algorithm program was written in Matlab (Ver. 7.0.4). In this paper, bee algorithm (BA) was used to select the best descriptors. Bee algorithm is a new population-based optimization algorithm, which is derived from the observation of real bees and proposed to feature selection [2]. Four descriptors were selected and used to model building by multiple linear regression (MLR) method. Root mean square error (RMSE) and determination coefficient (R2) were obtained as 0.157 and 0.9642, respectively. Then the data set was divided to training and test sets and the RMSE and R2 were obtained for two sets using four selected descriptors by BA. The RMSE and R2 were obtained as 0.148 and 0.9545 for the training set and 0.2630 and 0.9366 for the test set.