NetAllergen

NetAllergen predicts protein allergenicity by using a Random Forest machine learning model integrated with MHC class II presentation propensity to improve identification of IgE-inducing allergens.

Key Features:

• Random Forest Algorithm: Employs the Random Forest machine learning algorithm to model allergenicity, reducing overfitting and improving predictive accuracy.
• MHC Class II Presentation Propensity: Integrates MHC class II presentation propensity as a feature to enhance prediction beyond sequence similarity, improving detection when homology to known allergens is low.
• Comprehensive Allergen Dataset (AllergenOnline): Utilizes a curated dataset from AllergenOnline containing IgE-inducing allergens for model training.
• Protein Partitioning and Redundancy Removal: Applies a novel protein partitioning pipeline to remove redundancy within the dataset and ensure diverse, high-quality training data.
• Comparative Performance: Demonstrates improved prediction accuracy compared to sequence similarity methods such as BLAST and previous predictors like AlgPred 2.

Scientific Applications:

• Allergy Research: Aids identification and characterization of potential allergenic proteins for basic and translational allergy studies.
• Food Safety Assessment: Predicts potential allergenicity of novel food proteins to support risk assessment and development of hypoallergenic food sources.
• Public Health and Clinical Relevance: Supports efforts to reduce incidence of allergic reactions and to study atopic disorders in children and adults.

Methodology:

Allergens are sourced from the AllergenOnline curated database, redundancy is removed using a novel protein partitioning pipeline, and a Random Forest model is trained using features that include MHC class II presentation propensity.