Another major work of our group was submitted. We developed GraPPI, a structure-informed graph AI framework for learning protein–protein interactions. At its core is a self-supervised protein interface encoder. Designed as a unified and transferable PPI modeling framework, the learned embeddings can be adapted to diverse downstream tasks, including protein binding recognition, binding-affinity prediction, and mutation-effect analysis. Congratulations to Zhiyuan and the team on bringing this work together! Read more here.