Gemini quaternary ammonium surfactants (GQAS) have a unique structure built of two conventional surfactants connected by a spacer group. In previous studies, it has been found that GQAS have potency as antimicrobial agents. Thus, we developed a quantitative structure–activity relationship (QSAR) model to predict the antibacterial activity of GQAS. A dataset containing 57 GQAS with antibacterial activity against Escherichia coli was chosen from the literature. After optimizing all structures of these compounds using the ab initio 6-311G basis sets at the Hartree–Fock level theory, the molecular descriptors were calculated using the Mordred program. The genetic algorithm (GA) and multiple linear regressions (MLR) were used for generating two QSAR models with different splitting techniques. The predictive powers of the obtained models were discussed using the leave-one-out (LOO) cross-validation and external test set. The best GA-MLR models were obtained with reliable value of R2 = 0.891, Q2 LOO = 0.851, lack-of-fit = 0.116, root mean square error (RMSEtrain) = 0.267, R2 test = 0.834, and RMSEtest = 0.269. The GA-MLR methods were used to develop models that possess good predictive ability based on both internal and external validation parameters. The design of new molecules was done, and the antibacterial activity could be predicted using the resulting model with 16 compounds that showed potential as antibacterial agents.
Key words: Multiple linear regression, molecular descriptors, genetic algorithm, cationic surfactants, gemini quaternary ammonium, GA-MLR.
|
