A fuzzy neural network (FNN) was trained on a dataset of 177 HIV-1 protease ligands with experimentally measured IC(50) values. A set of descriptors was selected to build nonlinear quantitative structure-activity relationships. A genetic algorithm (GA) was implemented to optimize the architecture of the fuzzy neural network used to predict biological activity of HIV-1 protease inhibitors. Evolutionary methods were used to apply feature selection (FS) to this model. Results obtained on an external test set of 21 molecules, with and without feature selection, were compared. Applying feature selection to the GA-FNN resulted in a more accurate prediction of biological activity. Fuzzy IF/THEN rules were extracted from the optimized FNN. In the future the developed models are expected to be useful in the rational design of novel enzyme inhibitors for HIV-1 protease.