Background: Predicting the allergenicity of proteins is challenging. We considered the possibility that the properties of the intact protein that may alter the likelihood of being taken up by antigen presenting cells, may be useful adjuncts in predicting allergens and non-allergens in silico. It has been shown that negatively charged acidic proteins are preferentially processed by dendritic cells.
Methodology: Datasets (aeroallergen, food-allergen and non-allergen) for in-silico study were obtained from public databases. Isoelectric point (pI), net charge, and electrostatic potential (EP) were calculated from the protein sequence (for pI and net charge) or predicted structure (for EP).
Result: Allergens and non allergens differed significantly in pI, net charge, and EP (p<0.0001). Cluster analysis based on these parameters resulted in well defined clusters. Non-allergens were characterized by neutral to basic pI (mean+/-SE, 7.6+/-0.16) and positive charge. In contrast allergens were acidic (5.7+/-0.15) and negatively charged. Surface electrostatic potentials calculated from predicted structures were mostly negative for allergens and mostly positive for non-allergens. The classification accuracy for non-allergens was superior to that for allergens. Thus neutral to basic pI, positive charge, and positive electrostatic potentials characterize non-allergens, and seem rare in allergens (p<0.0001). It may be possible to predict reduced likelihood of allergenicity in such proteins, but this needs to be prospectively validated.