FoldEnsemble: A Multi-Modal Dataset and Model for Predicting Protein Structural Ensembles and Folding Kinetics

Assemble a standardized dataset linking structures to multi-modal dynamics data including sparse NMR restraints, paramagnetic NMR, HDX-MS, hydrogen–deuterium exchange protection patterns, and time-resolved experiments, combined with AF2 pLDDT/PAE maps and MD variability. Train an ensemble-predicting network to output conformer distributions and kinetic parameters (e.g., two-state vs. intermediate-rich), using Rosetta/NMR-guided modeling as a teacher. Prioritize computational efficiency borrowing parallel coordinate strategies from Cerebra. This reframes protein structure prediction from single static models to structure plus dynamics, using uncertainties as features rather than nuisances. The dataset serves as a community resource to evaluate ensemble prediction, addressing an acknowledged gap. It synthesizes Rosetta’s integration of sparse experimental data with modern deep learning predictors, extending beyond single-model accuracies to dynamics-aware metrics, and can cross-validate remote-homolog-inferred pathways from PAthreader. Predicting ensemble breadth and folding rates should improve mutation effect forecasts, design of conformational switches, and binder design to specific states. The impact is establishing the first broadly applicable benchmark and baseline models for dynamics prediction, catalyzing a shift from static to ensemble-centric protein modeling.

References:

1. Protein structure and folding pathway prediction based on remote homologs recognition using PAthreader. Kailong Zhao, Yuhao Xia, Fujin Zhang, Xiaogen Zhou, Stan Z. Li, Guijun Zhang (2023). Communications Biology.
2. Highly accurate protein structure prediction with AlphaFold. J. Jumper, Richard Evans, A. Pritzel, Tim Green, Michael Figurnov, O. Ronneberger, Kathryn Tunyasuvunakool, Russ Bates, Augustin Žídek, Anna Potapenko, Alex Bridgland, Clemens Meyer, Simon A A Kohl, Andy Ballard, A. Cowie, Bernardino Romera-Paredes, Stanislav Nikolov, Rishub Jain, J. Adler, T. Back, Stig Petersen, D. Reiman, Ellen Clancy, Michal Zielinski, Martin Steinegger, Michalina Pacholska, Tamas Berghammer, Sebastian Bodenstein, David Silver, O. Vinyals, A. Senior, K. Kavukcuoglu, Pushmeet Kohli, D. Hassabis (2021). Nature.
3. Recent Advances in NMR Protein Structure Prediction with ROSETTA. Julia Koehler Leman, Georg Künze (2023). International Journal of Molecular Sciences.
4. Cerebra: a computationally efficient framework for accurate protein structure prediction. Jian Hu, Weizhe Wang, Haipeng Gong (2024). bioRxiv.