Application of Image Processing for QSAR Study of Quinolone Derivatives as Anti-malaria Agents using different Chemometrics Methods

A quantitative structure activity relationship (QSAR) study was performed using image processing method to obtain a model that best describes the anti-malaria activities (EC50) of quinolone derivatives. In this study, pixel of molecular structures was considered as descriptors. Accordingly, a dataset of 28 quinolone derivatives was collected from literature [1]. Then structures were drawn in ChemOffice package and converted to pixels using MATLAB software. In order to have an appropriate selection of pixels, pre-processed methods such as mean centering and scaling were used. The useful descriptors were then selected using principal component analysis method (PCA) and genetic algorithm (GA) methods and the final model was developed by partial least squares (PLS) multivariate calibration method [2,3]. The QSAR model performance was evaluated and the agreement between computational and experimental values was investigated. The statistical parameters of all models were compared such as RMSEP (Best model, 0.0183), RSEP (Best Model, 0.246) and R2 (Best Model, 0.9986). The developed model showed high ability to predict anti-malaria activities (EC50).