In parallel we have developed a number of approaches based on information extraction to complement the description of interaction networks. In particular as part of BIOCREATVE II.5 challenge (http://www.biocreative.org) we have developed a meta-server that combines the information provided by text-mining servers about protein interactions described in scientific publications. The results of these systems are complementary to the ones provided by authors and database curators (Biocreative II.5 experiment on Structure Digital Abstracts, Leitner et al., 2010). During the presentation I will highlight the possibilities offered by this type of systems and how they can be integrated in systems accessible to experimental biologists and general users.
Based on the information on protein networks obtained with these and other methods we have proposed use of networks to connect disease associated genetic variants with the biomedical meta-information in the context of cancer genome projects. (Baudot et al., Genome Biology 2009). The result of the first analysis (Baudot et al., 2010) of common and specific functions associated to a collection of cancer types will be discussed.