Sepsis is deadly because it varies so greatly between different patients, and within individual patients over time, making it hard to diagnose and treat effectively. But scientists writing in Frontiers in Science explain how we can tackle the challenge by understanding sepsis as the result of multiple types of immune dysregulation. Using systems immunology to understand its mechanisms and target treatments, the researchers argue that we can tackle a major under-recognized cause of deaths worldwide, prevent deaths from future pandemic diseases, and treat survivors’ lingering symptoms.

Researchers from Zhejiang University and HKUST (Guangzhou) have introduced ProtET, an innovative AI-powered multi-modal protein editing model published in Health Data Science. ProtET leverages advanced transformer-structured encoders and a hierarchical training paradigm to align protein sequences with natural language instructions, enabling precise and controllable protein editing.

The model was trained on over 67 million protein-biotext pairs from Swiss-Prot and TrEMBL databases and demonstrated significant improvements across key benchmarks, including 16.9% enhanced protein stability, optimized enzyme catalytic activity, and improved antibody-antigen binding affinity. ProtET’s zero-shot capabilities successfully designed SARS-CoV antibodies with stable 3D structures, highlighting its real-world biomedical applications.

This research represents a major advancement in AI-driven protein engineering, offering a scalable and interactive tool for scientific discovery, synthetic biology, and therapeutic development.