A suite of tools to mine relations of nuclear receptors and cofactors from
scientific literature

Dietrich Schuhmann
LION Bioscience AG

At present information extraction in the biological field raises strong interest
since the biologist has to oversee large amounts of data, which comes from
scientific literature. The early approaches concentrated on protein-protein
interactions, since it seems to be that the correct information can be extracted
easily with the help of dictionaries which contain the names of the proteins.
This raises the question, what the correct information might be, and relates to
the evaluation of the result according to a professional biologist. We have
developed special software, which extracts information and offers results to the
biologist. Furthermore, we have chosen the topic 'nuclear receptors' to extract
the relevant information for the biologists in the laboratory. The results are
different software components, which support the information extraction process.
Amongst them are: (1) the core technology for information extraction, (2) a
clustering tool with a viewer, (3) a tool to retrieve information on a sentence
basis and on a co-occurrence basis, (4) a general relation viewer (pSCOUTTM),
and (5) a tool to curate part of the data, that we are looking at. From the
curation process attached to the last tool we generated a 'gold standard' of
example sentences, where other especially automatic techniques can be compared
to. One of the core issues in information extraction in molecular biology is the
definition, registration and the curation of the named entities.