NeuroPred is a web application designed to predict cleavage sites at basic amino acid locations in neuropeptide precursor sequences. The user can study one amino acid sequence or multiple sequences simultaneously, selecting from several prediction models and optional, user-defined functions. Logistic regression models are trained on experimentally verified or published cleavage data from mollusks, mammals and insects, and amino acid motifs reported to be associated with cleavage. Confidence interval limits of the probabilities of cleavage indicate the precision of the predictions; these predictions are transformed into cleavage or non-cleavage events according to user-defined thresholds. In addition to the precursor sequence, NeuroPred accepts user-specified cleavage information, providing model accuracy statistics based on observed and predicted cleavages. Neuropred also computes the mass of the predicted peptides, including user-selectable post-translational modifications. The resulting mass list aids the discovery and confirmation of new neuropeptides using mass spectrometry techniques. The NeuroPred application, manual, reference manuscripts and training sequences are available at http://neuroproteomics.scs.uiuc.edu/neuropred.html.