AI Faces Challenges in Modeling Dynamic Proteins like AGP

A study by researchers from Brussels reveals the limitations of AI tools such as AlphaFold in predicting flexible and dynamic regions of proteins like alpha-1-acid glycoprotein (AGP). The research highlights how small genetic mutations and glycosylation impact the protein's behavior, which AI models often oversimplify or miss entirely. While AlphaFold performs well for rigid parts, it struggles with the flexible, functionally important areas. The findings emphasize the need for experimental validation and cautious interpretation of AI predictions in biomedical research.