Automatic Extraction of Information on Protein-Protein Interactions from Biological Abstracts

Abstract

Due to the proliferation of readily available but overwhelming research results in molecular biology, there is an increasing need for automatic information extraction (IE) to support database building and to intelligently find novel knowledge in online journal collections. Extraction of substance names and other terms has gained success to a certain extent, and for recent years automated extracting of interactions between proteins have gained great attentions. Many of the previous researchers extracted information on protein interactions by matching the sentences in the biological literature with hand-tailored patterns in regular expressions on some pre-defined set of verbs representing a certain type of interaction. However, although pattern-matching based information extraction can be effective and quick on limited types of interactions in a limited domain, the workload of preparing patterns for extraction might be too expensive if we expand our attentions to other domains or to a wider scope of interaction types. In this talk, we first introduce a conventional pattern-matching method for extracting protein-protein interactions from the biological literature, and then an alternative bio-information extraction method based on full parsing with a large-scaled general-purpose grammar.