Researchers continued to build better AI-based models for protein structure prediction and design, notably Alphafold31 and RoseTTAFold All-Atom2. Last year’s versions could model how peptides could fold into their final shapes, but these new models can show how folded proteins bind and interact with other small molecules, including DNA, RNA and other proteins. As training data continue to grow and improve, these models will perform even better, leading to better prediction of protein functions and possible drug design.
