Protein-based drug discovery depends on understanding how molecules move and fold, but traditional molecular dynamics simulations are slow and resource-intensive, limiting how many protein variants researchers can realistically evaluate before wet-lab testing.
Astellas Pharma integrated BioEmu, a biomolecular simulation model from Microsoft Research, through Microsoft Foundry, building a custom Azure Machine Learning environment with an API-based interface so researchers could explore protein conformations without managing GPU infrastructure directly.
In a single proof of concept, Astellas evaluated hundreds of protein mutations in parallel — running 100 mutation-scan datasets simultaneously instead of sequentially — accelerating candidate prioritization with conformational structural insight and expanding how comprehensively teams can explore protein variant space.