QSPR analysis for melting point of fatty acids using genetic algorithm based multiple linear regression (GA-MLR)

Abstract

A quantitative structure–property relationship (QSPR) study was performed to correlate the melting point of fatty acids with their molecular structures. The entire set of 62 fatty acids was divided into a training set of 48 acids and a test set of 14 acids according to the DUPLEX algorithm. An extended set of molecular descriptors was calculated to represent the molecular structures by DRAGON software. Genetic algorithm based multiple linear regression (GA-MLR) was used to select descriptors and develop models. A five-descriptor equation was obtained for the training set, with a squared correlation coefficient (R2) of 0.947 and a standard error (s) of 9.3. The reliability of the proposed model was further validated through the test set. Satisfactory results of R2=0.976 and a mean absolute error of 5.7 for the test set confirmed the model being very useful to predict the melting point of fatty acids.