Application of genetic algorithms for pixel selection in multivariate image analysis for a QSPR study of half-wave potentials for benzoxazinescompounds

The Half-wave potential (E1/2), which is an important electrochemical property for a reversible oxidation-reduction system, can be useful for predicting other electrochemical properties and activities of organic compounds. In this study, a quantitative structure-property relationship (QSPR) analysis has been directed to a series of 40 benzoxazines compounds with half-wave potentials (E1/2) (1) property that was performed by chemometrics methods. Bidimensional images were used to calculate some pixels (2). Multivariate image analysis was applied to QSPR modeling of the E1/2 potential of benzoxazines derivatives by means of multivariate calibration such as principal component regression (PCR) and partial least squares (PLS) (3). In this paper we investigate the effect of pixel selection by application of genetic algorithms (GAs) for PLS model (4). GAs is very useful in the variable selection in modeling and calibrationbecause of the strong effect of the relationship between presence/absence of variables in a calibration model and the prediction ability of the model itself. The subset of pixels, which resulted in the low prediction error, was selected by genetic algorithm. The resulted model showed high prediction ability with RMSEP of 0.1428, 0.1105 and 0.0103 for PCR, PLS and GA-PLS models, respectively. Furthermore, the proposed QSPR model with GA-PLS can contribute to the E1/2, and can be useful in predicting the E1/2 of the other compounds

Keywords:

Half-wave potential , QSPR , multivariate image analysis , PCR , PLS , genetic algorithms