AlphaFold as a prior: experimental structure determination conditioned on a pretrained neural network
Open accessPublished version of a preprint
Abstract
Advances in machine learning have transformed structural biology, enabling swift and accurate prediction of protein structure from sequence. However, key challenges persist in modeling side-chain packing, condition-dependent conformational changes and biomolecular interactions, largely because of limited high-quality training data. At the same time, emerging experimental techniques such as cryo-electron microscopy (cryo-EM), cryo-electron tomography (cryo-ET) and high-throughput crystallography are generating vast amounts of structural information but converting these data into mechanistically interpretable atomic models often remains difficult. Here we show that integrating experimental measurements directly into protein structure prediction can overcome these limitations. We introduce ROCKET, an augmentation of AlphaFold2 that refines predicted structures using cryo-EM, cryo-ET and X-ray crystallography data. By optimizing structures in the space of coevolutionary embeddings rather than Cartesian coordinates, ROCKET captures biologically meaningful structural variation that is inaccessible to AlphaFold2 alone and to existing automated modeling approaches, especially when the signal-to-noise ratio is low. ROCKET enables scalable, automated model building without retraining and provides a general framework for integrating experimental observables with biomolecular machine learning.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
Highly accurate protein structure prediction with AlphaFold
Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A., et al. (2021) · Nature · doi:10.1038/s41586-021-03819-2 · Cited by 43,764 · cited 1× · see citing works
“Analogous to AMBER relaxation in standard AF2 pipelines 1 , this step further optimizes geometry (Supplementary Fig.”
Accurate structure prediction of biomolecular interactions with AlphaFold 3
Abramson, J., Adler, J., Dunger, J., Evans, R., Green, T., Pritzel, A., Ronneberger, O., Willmore, L., Ballard, A. J., Bambrick, J., et al. (2024) · Nature · doi:10.1038/s41586-024-07487-w · Cited by 15,861 · cited 1× · see citing works
“Despite this accomplishment, computational approaches still struggle to capture important properties such as side-chain packing, functional dynamics and large-scale molecular assembly 5–9 .”
Protein complex prediction with AlphaFold-Multimer
Evans, R., O’Neill, M., Pritzel, A., Antropova, N., Senior, A., Green, T., Žídek, A., Bates, R., Blackwell, S., Yim, J., et al. (2021) · doi:10.1101/2021.10.04.463034 · Cited by 3,808 · see citing works
Automated model building and protein identification in cryo-EM maps
Jamali, K., Käll, L., Zhang, R., Brown, A., Kimanius, D., & Scheres, S. H. W. (2023) · doi:10.1101/2023.05.16.541002 · Cited by 686 · see citing works
Predicting multiple conformations via sequence clustering and AlphaFold2
Wayment-Steele, H. K., Ojoawo, A., Otten, R., Apitz, J. M., Pitsawong, W., Hömberger, M., Ovchinnikov, S., Colwell, L., & Kern, D. (2024) · Nature · doi:10.1038/s41586-023-06832-9 · Cited by 545 · see citing works
Ahdritz, G., Bouatta, N., Floristean, C., Kadyan, S., Xia, Q., Gerecke, W., O’Donnell, T. J., Berenberg, D., Fisk, I., Zanichelli, N., et al. (2024) · Nature Methods · doi:10.1038/s41592-024-02272-z · Cited by 439 · cited 18× · see citing works
“) to maximize agreement between an OpenFold prediction ( x prediction ) and experimental data ( E e ).”
InfluentialSelf-citationHigh‐accuracy protein structure prediction in <scp>CASP14</scp>
Pereira, J., Simpkin, A. J., Hartmann, M. D., Rigden, D. J., Keegan, R. M., & Lupas, A. N. (2021) · Proteins: Structure, Function, and Bioinformatics · doi:10.1002/prot.26171 · Cited by 314 · cited 2× · see citing works
“Breakthrough models such as AlphaFold2 (AF2) 1 , RoseTTAFold 2 and their descend-ants harness coevolutionary signals in large-scale sequence data to produce predictions with atomic-level precision and near-experimental accuracy 3,4 .”
SPEACH_AF: Sampling protein ensembles and conformational heterogeneity with Alphafold2
Stein, R. A., & Mchaourab, H. S. (2022) · PLOS Computational Biology · doi:10.1371/journal.pcbi.1010483 · Cited by 258 · see citing works
High-throughput prediction of protein conformational distributions with subsampled AlphaFold2
Monteiro da Silva, G., Cui, J. Y., Dalgarno, D. C., Lisi, G. P., & Rubenstein, B. M. (2024) · Nature Communications · doi:10.1038/s41467-024-46715-9 · Cited by 196 · see citing works
Assessing the utility of <scp>CASP14</scp> models for molecular replacement
Millán, C., Keegan, R. M., Pereira, J., Sammito, M. D., Simpkin, A. J., McCoy, A. J., Lupas, A. N., Hartmann, M. D., Rigden, D. J., & Read, R. J. (2021) · Proteins: Structure, Function, and Bioinformatics · doi:10.1002/prot.26214 · Cited by 62 · see citing works
Self-citationZhang, Y., Zhang, Z., Kagaya, Y., Terashi, G., Zhao, B., Xiong, Y., & Kihara, D. (2025) · Communications Biology · doi:10.1038/s42003-025-08783-5 · Cited by 19 · cited 1× · see citing works
“Recent efforts to adapt ML-based structure prediction to incorporate experimental constraints have done so either through fine-tuning of weights from existing architectures 26,27 or full training of a novel architecture, as exemplified by the ModelAngelo method for cryo-EM data 28 .”
Flagged first, then by citations
- Authors
- Fadini, A., Li, M., McCoy, A. J., Banjara, S., Okumura, H., Napier, E., Fontana, P., Khan, A. R., Jovine, L., Terwilliger, T. C., Read, R. J., Hekstra, D. R., & AlQuraishi, M.
- Affiliations
- University of Cambridge (GB)
- Harvard University (US)
- New Mexico Consortium (US)
- St. Jude Children's Research Hospital (US)
- Trinity College Dublin (IE)
- Karolinska Institutet (SE)
- Columbia University (US)
- Corresponding author
- Read, Randy J.
- Hekstra, Doeke R.
- AlQuraishi, Mohammed
- ORCID
- Venue
- Nature Methods
- Published
- 2026-04
- Volume
- 23
- Issue
- 4
- Pages
- 785-795
- ISSN
- 1548-7091 (ISSN-L)
- 1548-7105
- PMC
- PMC11870471
- OpenAlex
- W4407820310
- Cited by
- 22
- References
- 11
- Citations by year
- 2026: 16 · 2025: 6
- FWCI
- 32.83
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Locations
- Nature Methods (publishedVersion, cc-by)
- bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- Apollo (University of Cambridge) (submittedVersion, cc-by)
- PubMed Central (submittedVersion, cc-by)
- Funders
- Wellcome Trust (Z/17/Z, 209407/Z/17/Z, 209407)
- Kempe Foundation
- Stiftelsen för Strategisk Forskning
- Knut och Alice Wallenbergs Stiftelse (2018.0042)
- Vetenskapsrådet (2024-05336, 2018.0042)
- Umeå Universitet
- Kempestiftelserna
- Science for Life Laboratory
- National Institutes of Health (1dp2gm141000-01, GM063210)
- Directorate for Biological Sciences
- Biotechnology and Biological Sciences Research Council (BB/Y009398/1)
- National Institute of General Medical Sciences (P01GM063210, DP2-GM141000, GM063210, R35GM150546)
- Topics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
- Physical Sciences › Materials Science › Materials Chemistry › Enzyme Structure and Function
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Machine Learning in Bioinformatics
- Keywords
- AlphaFold2, rockets, protein structure refinement, cryo-electron tomography, cryo-electron microscopy, X-ray crystallography, protein conformational dynamics, domain rearrangement, automatic model generation, multimeric proteins
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
