Generative AI screened hundreds of spike-protein mRNA sequences before a single one hit the lab
Moderna · 2020–ongoing
What they did
In its 2020 COVID-19 vaccine sprint, Moderna used generative sequence-design models to produce hundreds of candidate mRNA permutations for the SARS-CoV-2 spike protein, computationally pre-screening them so only the most promising went to wet-lab testing — collapsing a normally sequential design step into a parallel one.
What happened
The practice has since scaled into standing infrastructure: Moderna deployed ChatGPT Enterprise to ~3,000 staff and built an internal tool ('mChat') reaching 80% adoption, later expanding to 750+ internal custom GPTs including a clinical-data tool ('Dose ID') and a contract-summarization tool. It's also applying generative design to a personalized cancer vaccine program (mRNA-4157) and, with IBM, a foundation model (MoLFormer) for lipid-nanoparticle and mRNA optimization.
The so-what
The highest-leverage use of generative AI in drug development isn't a chatbot feature — it's using generation-plus-screening to explore a design space (sequences, formulations) too large to hand-search, before any wet lab time is spent.