Fig. 1

FamPlex links named entities to protein families and complexes and their constituents. a Structure of FamPlex content. The affixes in gene_prefixes.csv can be used to improve recognition of molecular entity names, which can be linked to database identifiers using the lexical synonyms in grounding_map.csv. FamPlex itself contains identifiers representing families and complexes which are mapped to corresponding identifiers in other databases in equivalences.csv. Hierarchical relationships among families, complexes, and genes are listed in relations.csv. b Workflow for curation and evaluation. A gene list was used to define a corpus of articles that was divided into two subsets, “training” and “test”. The “training” corpus was processed with REACH and results were evaluated and used to guide curation. The “test” corpus was processed after incorporation of FamPlex and results were compared against the baseline from the training corpus