Relationship Mining Among the Entities Associated with GPCRs

G-protein-coupled receptors (GPCRs) share a characteristic core composed of seven-transmembrane ¿-helices represent by far the largest family of cell-surface molecules involved in signal transmission, accounting for ¿2% of the total genes encoded by the human genome. At present, most data types supplied by databases of GPCRs and relative proteins are mainly protein structure data, ligand binding data, interaction data of interactions with G-proteins and effectors, mutation data, sequence-derived data and protein family list. However, interaction data describe relationships of GPCRs related drugs and diseases are rare. We construct a measurement of relationships between GPCRs relative information based on calculating distances between entities (genes, diseases and drugs) in sentences of abstracts which are selected by key words ¿human¿ and nomenclatures of G-protein-coupled receptor and its family members. By using integrated method of distance measurement and mutual rank for relationship extraction, we obtain 88165 relationships and evaluate the result by randomly selected 700 sentences with an F-score of 67.58% and recall of 62.82%.