Using machine learning techniques and genomic/proteomic information from known databases for PPI prediction

Abstract

In current Proteomics, prediction of protein-protein interactions (PPI) is a crucial aim as these interactions take part in most essential biological processes. In this paper, we propose a new approach to PPI dataset processing based on the extraction information from well-known databases and the application of data mining techniques. This approach will provide very accurate Support Vector Machine models, trained using high-confidence positive and negative examples. Finally, our proposed model has been validated using experimental, computational and literature-collected datasets. © 2011 Springer-Verlag Berlin Heidelberg.