Figure 1

Pipeline for LAITOR's development. LAITOR's has been evaluated for correct classification of abstracts relevant to curation of protein-protein interactions from BioCreative II challenge (evaluation dataset). Co-occurrences of terms from the human proteins dictionary in these abstracts have been used as an indicator of relevance. Precision and recall have been measured as 0.89 and 0.48 respectively for the type 1 of co-occurrences. Afterwards, for abstracts ranked to be related to host-pathogen interactions in plants (application dataset), LAITOR has generated a list of co-occurrences and a network representation of the terms from the plant protein dictionary which could be found in this dataset. Symbol key: dark rectangles: dictionaries; grey shapes: abstract datasets; cylinders: public databases; ellipse: LAITOR script; white rectangles: LAITOR's outputs;