Structure-toxicity modeling of pesticides to honey bees.

A quantitative structure-activity relationship (QSAR) model was derived for estimating the acute toxicity of pesticides on the honey bee. Chemicals were described by means of autocorrelation descriptors encoding lipophilicity (H), molar refractivity (MR) and the H-bonding acceptor ability (HBA) of the pesticides. A three-layer feedforward neural network trained by the back-propagation algorithm was used as statistical engine for deriving a powerful QSAR model. The root mean square residual (RMSR) values for the training and testing sets were 0.430 and 0.386, respectively. The practical interest of this original model was discussed.