Text mining for systems biology and MetNet.

摘要

The rapidly expanding volume of biological and biomedical literature motivates demand for more friendly access. Better automated mining of this literature can help find useful and desired citations and can extract new knowledge from the massive biological "literaturome." The research objectives presented here, when met, will provide comprehensive text mining utilities within the MetNet (Metabolic Network Exchange) (Wurtele et al., 2007), platform to help biologists visualize, explore, and analyze the biological literaturome. The overarching research question to be addressed is how to automatically extract biomolecular interactions from numerous biomedical texts. Here are the specific aims of this work. (1) Research on the text empirics of interaction-indicating terms to find more clues to improve the current algorithm applied in PathBinder to more precisely judge whether biomolecular interaction descriptions are present in sentences from the biological literature. (2) Based on these research results, extract interacting biomolecule pairs from literature and use those pairs to construct a biomolecule interaction database and network. (3) Integrate biomolecular interaction-indicating term extraction into MetNet's existing metabolomic network database. (4) Apply all of the above results in PathBinder software. (5) Quantitatively evaluate the success of algorithms developed based on the text empirics results. This work is expected to advance systems biology by answering scientific questions about biological text empirics, by contributing to the engineering task of building MetNet and key constituent subsystems of MetNet, and by supporting the MetNet project through selected maintenance tasks.