Bioinformatic Methods for Allergenicity Assessment Using a Comprehensive Allergen Database

<i>Background:</i> A principal aim of the safety assessment of genetically modified crops is to prevent the introduction of known or clinically cross-reactive allergens. Current bioinformatic tools and a database of allergens and gliadins were tested for the ability to identify potential allergens by analyzing 6 <i>Bacillus thuringiensis</i> insecticidal proteins, 3 common non-allergenic food proteins and 50 randomly selected corn <i>(Zea mays)</i> proteins. <i>Methods:</i> Protein sequences were compared to allergens using the FASTA algorithm and by searching for matches of 6, 7 or 8 contiguous identical amino acids. <i>Results:</i> No significant sequence similarities or matches of 8 contiguous amino acids were found with the <i>B. thuringiensis</i> or food proteins. Surprisingly, 41 of 50 corn proteins matched at least one allergen with 6 contiguous identical amino acids. Only 7 of 50 corn proteins matched an allergen with 8 contiguous identical amino acids. When assessed for overall structural similarity to allergens, these 7 plus 2 additional corn proteins shared ≧35% identity in an overlap of ≧80 amino acids, but only 6 of the 7 were similar across the length of the protein, or shared >50% identity to an allergen. <i>Conclusions:</i> An evaluation of a protein by the FASTA algorithm is the most predictive of a clinically relevant cross-reactive allergen. An additional search for matches of 8 amino acids may provide an added margin of safety when assessing the potential allergenicity of a protein, but a search with a 6-amino-acid window produces many random, irrelevant matches.