ABSTRACT
Nonlinear quantitative structure–activity relationship (QSAR) models were derived for predicting the adulticidal activity of piperidines against mosquitoes. Thirty-three chemicals with their 24-hour LD50 values, obtained under the same experimental conditions on Aedes aegypti females, were used. Energy optimization allowed us to consider two conformational series of structures that were described by means of CODESSA descriptors. Only those encoding physicochemical information were considered. The 1-octanol/water partition coefficient (log P) of the molecules was also computed from the AUTOLOGP™ software. Each series was subject to modeling from a three-layer perceptron (TLP). The models were derived from training sets including 70% of the molecules, which were randomly selected. The remaining molecules were used as external test sets. Only 3/2/1 TLP models were considered. A consensus model was computed with an interesting configuration coming from each series. The selected hybrid TLP model showed correlation coefficients of 0.93 and 0.97 on 298the training set and external test set, respectively. From the simulation results, various molecules were identified and searched for their availability in existing free databases. A list of molecules of potential interest was elaborated that will be further refined in order to find new molecules having a synergistic activity with pyrethroids, especially deltamethrin.
