Prediction of protein allergenicity using local description of amino acid sequence.

: The constant increase in atopic allergy and other hypersensitivity reactions has intensified the need for successful therapeutic approaches. Existing bioinformatic tools for predicting allergenic potential are primarily based on sequence similarity searches along the entire protein sequence and do not address the dual issues of conformational and overlapping B-cell epitope recognition sites. In this study, we report AllerPred, a computational system that is capable of capturing multiple overlapping continuous and discontinuous B-cell epitope binding patterns in allergenic proteins using SVM as its prediction engine. A novel representation of local protein sequence descriptors enables the system to model multiple overlapping continuous and discontinuous B-cell epitope binding patterns within a protein sequence. The model was rigorously trained and tested using 669 IUIS allergens and 1237 non-allergens. Testing results showed that the area under the receiver operating curve (AROC) of SVM models is 0.81 with 76 percent sensitivity at specificity of 76 percent . This approach consistently outperforms existing allergenicity prediction systems using a standardized testing dataset of experimentally validated allergens and non-allergen sequences.