Figure 2.

The text mining approach and information integration for TCMGeneDIT. Literature corpus about TCMs was collected from PubMed and used for entity annotations and information extraction. Annotated documents were mined based on hypothesis test and collocation analysis to discover entity associations. On the other hand, the raw corpus was pre-processed with public tools and several rules can be applied to the processed sentences for extracting the relations between effecters and effects. The constructed effect set was used to filter candidate relation and literature annotation. Protein-protein interactions and biological pathways were integrated into our database and disease candidate genes from PharmGKB were used for transitive inference. Users can access the database via the web interface. Thick arrow indicates the basic work flow of TCMGeneDIT.