Abstract: Predictions of the binding ability of antigen peptides to Major Histocompatibility Complex（MHC） class II molecules are important for immunology research and vaccine design.The variable length and other aspects of each binding peptide complicate this prediction.In this paper，generalized selective neural network ensemble is proposed for prediction of MHC class II-binding peptides，the ensemble is built on two-level ensemble architecture.The first-level ensemble is used to create primary Neural Network Ensemble（NNE），where differential evolution is used to build some NNEs.The second-level ensemble is that a subset of primary NNEs is selected to make up the final ensemble.Experiment results indicate that the generalized ensemble model has better generalization and performance compared to traditional selective neural network ensemble.