Biomedical text mining for disease-gene discovery

Abstract

Motivation and Objectives Because of the amount of electronic literature now available, it is challenging for biologists to search biomedical corpuses for any kind of desired information beyond simple text retrieval. Several tools have been developed to make text mining easier for them. Some of these tools focus on extracting biomedical terms; such as protein names and biological processes, given any input text. The tools COREMINE Medical (http://www. coremine.com, last accessed on 25 September 2012) and GoPubMed: http://www.gopubmed. com, last accessed on 25 September 2012) are just two examples. Other tools apply rule-based strategies to relate biomedical concepts to each other. E.g., BITOLA (http://ibmi.mf.uni-lj.si/bitola/ last accessed on 25t September 2012) (Hristovski et al., 2005). We have been developing a methodology and tool to discover genes implicated in any given disease or disorder. In fact, our tool takes from the user any free text query as an input and attempts to identify those genes most strongly linked to the query. As an output, the tool returns an ordered list of the best genes matching the query. The core work of our tool is based on text mining. Basically, each gene is linked to a profile that contains the biological terms that are most significant for it. Similarly, we link the input query to a corresponding keyword profile. The genes appearing at the top of the output list are the ones whose profiles are highly similar to that of the input query.