Applying Stacked and Cascade Generalizations to B-cell Epitope Prediction

Abstract

One of the major challenges in the field of vaccine design is to identify the B-cell epitopes in ever-evolving viruses. Various prediction servers have been developed to predict linear or conformational epitopes, each relying on different physicochemical properties and adopting distinct search strategies. We propose meta learning approaches to epitope prediction based on stacked generalization and cascade generalization. By combining the base prediction servers in a hierarchical architecture, we demonstrated that a meta learner outperformed the best single server in predicting the epitopes of an independent dataset of pathogen proteins.