Predicting cancer interaction networks using text-mining and structure understanding.

Extended biomolecule binding or interaction networks can be built by computationally predicting protein-protein interactions from diverse data sources. To construct networks focused on cancer research our approach combines domain specific natural language processing (NLP) assisted text-mining of biomedical literature databases with structure-based protein-protein interaction prediction reinforced with sub-cellular localization and evolutionary information. Fast retrieval of structure-based queries will be accomplished by using a novel knowledge discovery process developed previously.