ECCB 2024: Advancing Vaccine Development through Precise AI-driven Prediction of Protective Antigens

Antimicrobial resistance (AMR) is on the rise and projected to cause up to 10 million deaths yearly by 2050 (Tang et al. 2023, Br J Biomed Sci). This underscores the urgent medical need for new and effective therapies, like antimicrobial vaccines, to tackle the resistant pathogens. Traditional vaccine development approaches, such as Reverse Vaccinology, is time- consuming and resource-intensive, highlighting the demand for a more efficient target discovery. To address this, we developed EDENTM, an AI model for fast and precise identification of broadly protective antigens. EDENTM has proof-of-concept for bacterial vaccine development (Gulati et al. 2023, mBio) and is applied in ongoing vaccine programs targeting WHO prioritized AMR threats to global health. We here present the improved next generation of EDENTM which leverages state-of-the-art deep learning methodologies. First, a deep transformer model was trained for comprehensive feature representation of protein sequences based on a dataset of more than 300,000 annotated proteins. This deep transformer was then applied to encode features for training the EDENTM model to predict protective antigens. The training set of protective antigens was curated by screening scientific literature with Large Language Models using engineered prompts followed by manual confirmation by field experts. The resulting improved EDENTM model achieved state-of-the-art performance in predicting protective antigens and can be used for efficient vaccine development to prevent AMR infections.

Download poster