ALLERDET: A novel web app for prediction of protein allergenicity

Abstract

Allergic diseases are increasing around the world with unprecedented complexity and severity. One of the reasons is that genetically modified crops produce new potentially allergenic proteins. From this starting point, many researchers have paid attention to the development of tools to predict the allergenicity of new proteins. In this study, a novel approach is introduced for the prediction of food allergens based on Artificial Intelligence techniques: a pairwise sequence alignment with the FASTA program for feature extraction and the use of the Deep Learning technique known as Restricted Boltzmann Machines in combination with the Decision Tree method for the prediction process. The developed tool, called ALLERDET (publicly available at http://allerdet.frangam.com), overcomes the state-of-the-art methods. The performance of our method is: 98.46% sensitivity, 94.37% specificity and 97.26% accuracy), on a data set built from several publicly available sources.

Keywords: ALLERDET; Allergen detection; FASTA; Food allergy; Pairwise sequence alignment; Restricted Boltzmann Machines.