ProMiner: Organism-specific protein name detection using approximate string matching

were required, (2) disambiguation failed because of missing synonyms, e.g. ”vertebrate” and (3) for several cases the provided gold standard might be incorrect as considered abstracts describe findings in rat or human instead of mouse. Description Examples Unspecific synonym growth retarded, perinatal lethality, long lived Wrong context TGF-beta superfamily, c-myc tumors Unknown ambiguity high dose set at MTD or MFD Doubtful gold-standard interleukin-2 , H-2 locus, c-Jun Table 2: Sample of false positive matches in the best submitted mouse search run. Detected matches are marked . The first thirty false positive matches of the best submitted results for mouse (run number 3) have been analyzed in more detail. Undetected ambiguities are the dominant reasons for false positive matches (60%). This number includes unspecific synonyms which have neither been detected during curation nor marked as questionable synonyms, detection of synonyms in wrong contexts, and unknown external ambiguities. Detection of genes from other organisms accounts for 13.3 % of false matches. In eight cases, the reason for exclusion from the gold-standard remained unclear. Concluding, the ProMiner method provides high sensitivity using approximate matching and retains high specificity due to sensible incorporation of questionable synonyms.