Finding small molecule and protein pairs in scientific literature using a bootstrapping method.

The relationship between small molecules and proteins has attracted attention from the biomedical research community. In this paper a text mining method of extracting small-molecule and protein pairs from natural text is presented, based on a semi-supervised machine learning approach. The technique has been applied to the complete collection of MEDLINE abstracts and pairs were extracted and evaluated. The results show the feasibility of the bootstrapping system, which will subsequently be further investigated and improved.