Cascaded multiple classifiers for secondary structure prediction

We describe a neu classifier for protein secondary structure prediction tha t is formed by cascading together different types of classifiers using neur al networks and linear discrimination. The new classifier achieves an accur acy of 76.7% (assessed by a rigorous full Jack-knife procedure) on a new no nredundant dataset of 496 nonhomologous sequences (obtained from G.J. Barto n and J.A. Cuff). This database was especially designed to train and test p rotein secondary structure prediction methods, and it uses a more stringent definition of homologous sequence than in previous studies. We show that i t is possible to design classifiers that can highly discriminate the three classes (H, E, C) with an accuracy of up to 78% for beta-strands, using onl y a local window and resampling techniques. This indicates that the importa nce ut long-range interactions for the prediction of beta-strands has been probably previously overestimated.