Researchers Demand Explainable AI for Protein Design

Scientists from the Centre for Genomic Regulation are calling for the integration of explainable AI (XAI) into protein language models ^[1]. According to a perspective paper in Nature Machine Intelligence, current systems often operate as "black boxes" ^[1]. These models are increasingly used to design new enzymes for carbon capture and industrial catalysts ^[1][2].

Researchers emphasize that transparency is necessary to ensure AI-generated designs are safe and unbiased ^[1][3]. Without interpretability, scientists struggle to understand the biological principles behind AI decisions ^[1]. Moving toward open models could help researchers learn new rules of protein folding and function ^[3].