The text mining approach and information integration for MeInfoText. Literature about human, methylation and cancer is collected from PubMed and annotated with gene symbols. The most recent 100 gene-annotated abstracts are manually checked to reduce false named entity recognitions and enhance dictionary coverage. The gene-annotated documents are indexed with Plucene module and then mined according to the frequencies of co-occurrences of entities. Various association, protein-protein interaction and pathway information are stored in the relational database, MeInfoText. Users can search the database via the web interface. Thick arrow indicates the basic workflow of MeInfoText.