Simple ensemble methods are competitive with state-of-the-art data integration methods for gene function prediction

Several works showed that biomolecular data integration is a key issue to improve the prediction of gene functions. Quite surprisingly only little attention has been devoted to data integration for gene function prediction through ensemble methods. In this work we show that relatively simple ensemble methods are competitive and in some cases are also able to outperform state-of-the-art data integration techniques for gene function prediction.