Seminars

Understanding how tumor signaling changes during T cell attack may reveal targetable vulnerabilities. As T cell recognition is antigen-specific, a fundamental distinction arises between tumor cells bearing the target antigen and antigen-negative bystanders, yet whether their signaling responses differ remains unclear. Here, we implement a cell-type-specific phosphoproteomics workflow to profile these responses over 72 h of co-culture, providing a modular approach to resolving population-specific signaling perturbations in vitro.

Natural protein families contain vast functional diversity. Yet most protein engineering campaigns begin with a single parent sequence. To move beyond local optimization, we develop an active learning framework combining pLM embeddings with iterative model-guided selection to identify homologs with enhanced activity. Retrospective benchmarks on six experimental homolog landscapes show enrichment of top performers. We apply this approach to three industrially relevant enzyme families, enabling scalable discovery of highly active homologs.