Recent advances like AlphaFold3 have made protein structure prediction remarkably accurate. BioEmu moves forward beyond static snapshots to capture the dynamic nature of proteins essential to their function. By generating thousands of diverse conformations, BioEmu effectively emulates a protein’s equilibrium ensemble.
Trained on molecular dynamics trajectories, static crystal structures, and experimental stability data, BioEmu accurately models key functional motions—such as cryptic pocket formation, local unfolding, and domain rearrangements—and predicts relative free energies within ~1 kcal/mol of millisecond-scale MD simulations and experimental results. By integrating structural ensembles with thermodynamic predictions, BioEmu offers mechanistic insight and a scalable platform for understanding and engineering protein function.
BioEmu generates diverse protein conformations that capture intrinsic flexibility and dynamics, providing deeper insights into functional mechanisms. This approach improves functional annotation and supports applications such as drug design, protein engineering, and understanding disease-related mutations. By bridging static structure prediction with dynamic characterization, BioEmu highlights the importance of ensemble-based methods in protein biology.
More reading
Lewis, S., Hempel, T., Jiménez-Luna, J., Gastegger, M., Xie, Y., Foong, A. Y. K., García Satorras, V., Abdin, O., Veeling, B. S., Zaporozhets, I., Chen, Y., Yang, S., Foster, A. E., Schneuing, A., Nigam, J., Barbero, F., Stimper, V., Campbell, A., Yim, J., Lienen, M., Shi, Y., Zheng, S., Schulz, H., Munir, U., Sordillo, R., Tomioka, R., Clementi, C., & Noé, F. (2025). Scalable emulation of protein equilibrium ensembles with generative deep learning. Science, 373(6550), eadv9817. https://doi.org/10.1126/
SHARE
