Direct discovery · Biology · Medicine · Computer science
Virtual Lab designs experimentally active SARS-CoV-2 nanobodies
Experimentally characterized SARS-CoV-2 nanobodies, including two with improved binding to recent JN.1 or KP.3 variants while retaining ancestral-spike binding.
Summary
Swanson and colleagues describe a human-supervised Virtual Lab in which an LLM principal-investigator agent coordinates specialist agents to create and test a nanobody-design workflow. The resulting pipeline generated 92 SARS-CoV-2 nanobodies; experimental assays identified functional binders, including two with improved binding to recent viral variants.
AI role
An LLM-led multi-agent Virtual Lab coordinated a computational nanobody-design pipeline using ESM, AlphaFold-Multimer, and Rosetta.
Narrative role
This is direct-output evidence that an LLM-agent collaboration can contribute to a laboratory-validated biological design result, rather than only proposing hypotheses or evaluating research tasks.
Caveat
The project used substantial human direction and established computational biology tools; binding assays are an early validation step and do not establish therapeutic effectiveness.