Study of quantitative structure–property relationship for predicting the logP of pyrethroid derivatives using multiple linear regression method

In this research, predicting the logP of 34 types different pyrethroid derivatives was studied using quantitative structure-property relationship. The logP of studied pyrethroids was modeled using the genetic algorithm based on linear regression method (GA-MLR). It was found that the three effective descriptors GATS4P, PW3 and ZM1V have a reasonable correlation with logP, and led to the creation of a model with the highest regression coefficient and the lower error. The evaluation of GA-MLR model predictive ability for test set was performed by statistical parameters such as R2= 0.862, R2adj = 0.848, F=62.296 and MSE = 0.503. Also, the value of Q2LOO= 0.861 using the cross-validation method, and the values of R2 =0.880 and 0.929 for the training and test sets respectively, in the external validation method showed a very good correlation between experimental and prediction values. It was specified that the MLR model was reliable for predicting the logP of pyrethroid insecticides, and had sufficient accuracy with the lowest error.