In silico allergen identification: Proposal for a revision of FAO/WHO guidelines
Allergy is a widespread, often severe health problem. In vivo or in vitro identification of new allergenic proteins (natural or bioengineered) is time- and resource-consuming, and in vivo testing can be dangerous. Thus, allergenicity prediction through computation (in silico) was proposed to narrow down the number of potential allergens to be tested with traditional methods. In 2001, the Food and Agriculture Organization (FAO) and the World Health Organization (WHO) officially defined guidelines for in silico allergenicity prediction, based on amino acid sequence similarity to known allergens; these guidelines, however, have been criticized because of frequent false positives. In the present work, the BLAST (Basic Local Alignment Search Tool) software was used to compare known and potential allergens, and select only statistically significant homologies (i.e. homologies whose E value, calculated by BLAST, was < 1); FAO/WHO rules were then applied to these homologies. With this method, correct recognition of all known allergens, with only 10 false positives (1.26% of all predicted allergens) was achieved when using an upper limit of 0.1 for E values; complete suppression of wrong predictions, while maintaining 100% sensitivity, was obtained with little modifications of the minimum requirements contained in the FAO/WHO guidelines.
This work is licensed under a Creative Commons Attribution 3.0 License.