Inferring hidden relationships from biological literature with multi-level context terms

The Swanson's ABC model is powerful to infer hidden relationships buried in biological literatures. However, the model is inadequate to infer the relations with context information. In addition, the model generates very large amount of candidates from biological text, and it is the semi-automatic, labor intensive technique requiring human expert's input. In this paper, we propose a novel interaction inference technique that incorporates context term vectors into the ABC model to discover meaningful, hidden relationships. Our hypothesis is that the context-based relation extraction between AB interactions and BC interactions is more effective and efficient than the original ABC model without considering the context information. We evaluated our hypothesis with the datasets of the "Alzheimer's disease" related 77,711 PubMed abstracts. As golden standards, PharmGKB and CTD databases were used. The results indicate that context-based interaction extraction achieved better precision than the basic ABC model approach. The literature analysis also shows that interactions inferred by the context-based approach are more meaningful than interactions by the basic ABC model.