Enhancing Allergy Prediction Accuracy Through Machine Learning and ProteinBERT

Food allergies pose significant health risks globally, affecting a substantial portion of the population and necessi-tating more reliable diagnostic methods. As consumer demand drives modifications in raw materials, the likelihood of protein allergens in processed foods increases, highlighting the need for advanced detection techniques. This research explores leveraging Machine learning (ML) and Deep Learning(D L)to enhance allergy diagnosis accuracy and efficiency by analyzing the char-acteristics of amino acids. In this study, We employed both classical ML algorithms, such as K-Nearest Neighbors (KNN) and Random Forest, and DL models, incorporating advanced feature representations through ProteinBERT embeddings. Our findings reveal that the inclusion of ProteinBERT embeddings significantly improves model performance, with algorithms like KNN and Voting Classifier achieving accuracies over 90%. This approach not only enhances the precision of allergy prediction but also underscores the potential of integrating molecular-level data with machine learning techniques.