[Major revision of the allergen database for food safety (ADFS) and validation of the motif-based allergenicity prediction tool].

We have been maintaining an integral web server system, the Allergen Database for Food Safety (ADFS), since 2005 (http://allergen.nihs.go.jp/ADFS/). Recently, a group at the University of Nebraska-Lincoln released a new version of an allergen database, AllergenOnline. This database includes more than 1,300 allergens, all of which have been peer-reviewed by an international board of allergology experts. Here, we have totally revised the dataset of the ADFS by comparing it with that of AllergenOnline to improve the reliability of our allergen data. Moreover, the performance of our web-based tool for predicting new allergens (motif-based method), which was developed according to a theory proposed by Stadler & Stadler (2003), was validated using three methods. As a result of the integration of this allergen data, the number of (iso)allergens in the ADFS has increased to 1340, and epitope information is now available for 76 allergens. Using model datasets, the precision, recall, and specificity of our motif-based allergenicity prediction tool was proved to be 100.0%, 99.4%, and 100.0%, respectively. These results were similar to those for the original motif-based prediction model that was previously reported and are much better than those of the method recommended by FAO/WHO, especially with regard to the precision of predictions.