Learning millisecond protein dynamics from what is missing in NMR spectra

Many proteins’ biological functions rely on interconversions between multiple conformations occurring at micro-to millisecond (µs-ms) timescales. A lack of standardized, large-scale experimental data has hindered obtaining a more predictive understanding of these motions. After curating >100 Nuclear Magnetic Resonance (NMR) relaxation datasets, we realized an observable for µs-ms motion was hiding in plain sight. Millisecond motions can cause NMR signals to broaden beyond detection, leaving some residues not assigned in the chemical shift datasets of ∼10,000 proteins deposited in the Biological Magnetic Resonance Data Bank (BMRB)¹. We made the bold assumption that residues missing assignments are exchange-broadened due to µs-ms motions, and trained various deep learning models to predict missing assignments as markers for such dynamics. Strikingly, these models also predict µs-ms motion directly measured in NMR relaxation experiments. The best of these models, which we named Dyna-1, leverages an intermediate layer of the multimodal language model ESM-3². Notably, dynamics directly linked to biological function — including enzyme catalysis and ligand binding — are particularly well predicted by Dyna-1, which parallels our findings that residues with µs-ms motions are highly conserved. We anticipate the datasets and models presented here will be transformative in unlocking the common language of dynamics and function.