SHASI-ML: A Machine Learning-Based approach for immunogenicity prediction in Salmonella vaccine development

Ottavia Spiga ^* Anna Visibelli Francesco Pettini Bianca Roncaglia Annalisa Santucci

• Department of Biotechnology, Chemistry and Pharmacy, University of Siena, Siena, Italy

Accurate prediction of immunogenic proteins is critical for vaccine development and understanding host-pathogen interactions. This study presents SHASI-ML, a machine learning-based framework designed to predict immunogenic proteins in bacterial pathogens, specifically focused on Salmonella species. From a published dataset of immunogenic and non-immunogenic proteins, three groups of features were extracted from sequences and used to train and test the model, obtaining an outstanding ability to identify bacterial immunogens. SHASI-ML identified 292 immunogenic proteins, demonstrating an 89.3% detection rate for validated immunogens. The method achieved precision and specificity of 89.3% and 91.2%, respectively, with superior recall and F1-scores compared to existing approaches. Additionally, the importance of each group of features in the prediction was performed. These findings validate SHASI-ML as an efficient tool for prioritizing immunogenic candidates, reducing the time and cost of vaccine research. The results obtained in this work could be used to guide and optimize the research and industrialization of Salmonella vaccines.