AlphaFold2 fails to predict protein fold switching
Open accessPublished version of a preprint
Abstract
AlphaFold2 has revolutionized protein structure prediction by leveraging sequence information to rapidly model protein folds with atomic-level accuracy. Nevertheless, previous work has shown that these predictions tend to be inaccurate for structurally heterogeneous proteins. To systematically assess factors that contribute to this inaccuracy, we tested AlphaFold2's performance on 98-fold-switching proteins, which assume at least two distinct-yet-stable secondary and tertiary structures. Topological similarities were quantified between five predicted and two experimentally determined structures of each fold-switching protein. Overall, 94% of AlphaFold2 predictions captured one experimentally determined conformation but not the other. Despite these biased results, AlphaFold2's estimated confidences were moderate-to-high for 74% of fold-switching residues, a result that contrasts with overall low confidences for intrinsically disordered proteins, which are also structurally heterogeneous. To investigate factors contributing to this disparity, we quantified sequence variation within the multiple sequence alignments used to generate AlphaFold2's predictions of fold-switching and intrinsically disordered proteins. Unlike intrinsically disordered regions, whose sequence alignments show low conservation, fold-switching regions had conservation rates statistically similar to canonical single-fold proteins. Furthermore, intrinsically disordered regions had systematically lower prediction confidences than either fold-switching or single-fold proteins, regardless of sequence conservation. AlphaFold2's high prediction confidences for fold switchers indicate that it uses sophisticated pattern recognition to search for one most probable conformer rather than protein biophysics to model a protein's structural ensemble. Thus, it is not surprising that its predictions often fail for proteins whose properties are not fully apparent from solved protein structures. Our results emphasize the need to look at protein structure as an ensemble and suggest that systematic examination of fold-switching sequences may reveal propensities for multiple stable secondary and tertiary structures.
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.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The standard AlphaFold implementation detects only a single structural snapshot, the ground state, likely because its training set lacks a large set of redundant protein conformers.
“The standard implementation of AlphaFold performs well only in detecting a single structural snapshot (the “ground state” structure), likely due to the lack of a large set of redundant protein conformers in the training set16,32.”
From The power and pitfalls of AlphaFold2 for structure prediction beyond rigid globular proteins (2024) · cites this paper as [16]
AF2 has been demonstrated to fail in predicting multiple structures of metamorphic proteins, proteins with apo/holo conformational changes and other multi-state proteins using its default settings.
“yet AF2 has been demonstrated to fail in predicting multiple structures of metamorphic proteins11, proteins with apo/holo conformational changes12 and other multi-state proteins13 using its default settings.”
From Predicting multiple conformations via sequence clustering and AlphaFold2 (2024) · cites this paper as [11]
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 · see citing works
Matplotlib: a 2D graphics environment
Hunter, J. D. (2007) · Computing in Science & Engineering · doi:10.1109/mcse.2007.55 · Cited by 39,249 · see citing works
SciPy 1.0: fundamental algorithms for scientific computing in Python
Virtanen, P., Gommers, R., Oliphant, T. E., Haberland, M., Reddy, T., Cournapeau, D., Burovski, E., Peterson, P., Weckesser, W., Bright, J., et al. (2020) · Nature Methods · doi:10.1038/s41592-019-0686-2 · Cited by 38,430 · see citing works
seaborn: statistical data visualization
Waskom, M. (2021) · Journal of Open Source Software · doi:10.21105/joss.03021 · Cited by 7,310 · see citing works
Highly accurate protein structure prediction for the human proteome
Tunyasuvunakool, K., Adler, J., Wu, Z., Green, T., Zielinski, M., Žídek, A., Bridgland, A., Cowie, A., Meyer, C., Laydon, A., et al. (2021) · Nature · doi:10.1038/s41586-021-03828-1 · Cited by 3,025 · cited 4× · see citing works
“2 7 Although AlphaFold2 can sometimes be used to predict alternative quaternary structures, 16,39 here we show that it consistently fails to predict the conformational diversity of proteins that assume multiple secondary structure configurations, known as fold switchers.”
BackgroundMethodInfluentialDill, K. A., Ozkan, S. B., Shell, M. S., & Weikl, T. R. (2008) · Annual Review of Biophysics · doi:10.1146/annurev.biophys.37.092707.153558 · Cited by 1,163 · see citing works
AlphaFold and Implications for Intrinsically Disordered Proteins
Ruff, K. M., & Pappu, R. V. (2021) · Journal of Molecular Biology · doi:10.1016/j.jmb.2021.167208 · Cited by 635 · see citing works
Sampling alternative conformational states of transporters and receptors with AlphaFold2
del Alamo, D., Sala, D., Mchaourab, H. S., & Meiler, J. (2022) · eLife · doi:10.7554/elife.75751 · Cited by 511 · see citing works
Current structure predictors are not learning the physics of protein folding
Outeiral, C., Nissley, D. A., & Deane, C. M. (2022) · Bioinformatics · doi:10.1093/bioinformatics/btab881 · Cited by 82 · see citing works
Flagged first, then by citations
Research works printed but not linked: 34
- Lupas AN, Pereira J, Alva V, Merino F, Coles M, Hartmann MD. The breakthrough in protein structure predic- tion. Biochem J. 2021;478(10):1885-1890.
- Rose GD. Reframing the protein folding problem: Entropy as organizer. Biochemistry. 2021;60(49):3753-3761.
- David A, Islam S, Tankhilevich E, Sternberg MJE. The alphafold database of protein structures: A biologist's guide. J Mol Biol. 2021;434(2):167336.
- Murzin AG. Biochemistry. Metamorphic proteins. Science. 2008;320(5884):1725-1726.
- Porter LL, Looger LL. Extant fold-switching proteins are wide- spread. Proc Natl Acad Sci U S A. 2018;115(23):5968-5973.
- Tuinstra RL, Peterson FC, Kutlesa S, Elgin ES, Kron MA, Volkman BF. Interconversion between two unrelated protein folds in the lymphotactin native state. Proc Natl Acad Sci U S A. 2008;105(13):5057-5062.
- Bryan PN, Orban J. Proteins that switch folds. Curr Opin Struct Biol. 2010;20(4):482-488.
- Mishra S, Looger LL, Porter LL. A sequence-based method for predicting extant fold switchers that undergo alpha-helix $ beta-strand transitions. Biopolymers. 2021;112(10):e23471.
- Kabsch W, Sander C. Dictionary of protein secondary structure: Pattern recognition of hydrogen-bonded and geometrical fea- tures. Biopolymers. 1983;22(12):2577-2637.
- Fitzkee NC, Fleming PJ, Rose GD. The protein coil library: A structural database of nonhelix, nonstrand fragments derived from the pdb. Proteins. 2005;58(4):852-854.
- Quaglia F, Meszaros B, Salladini E, et al. Disprot in 2022: Improved quality and accessibility of protein intrinsic disorder annotation. Nucleic Acids Res. 2021;50:D480-D487.
- Zhang Y, Skolnick J. Tm-align: A protein structure alignment algorithm based on the tm-score. Nucleic Acids Res. 2005; 33(7):2302-2309.
- Xu J, Zhang Y. How significant is a protein structure similarity with tm-score=0.5? Bioinformatics. 2010;26(7):889-895.
- Pedregosa F, Varoquaux G, Gramfort A, et al. Scikit-learn: Machine learning in python. J Mach Learn Res. 2011;12:2825-2830.
- Mishra S, Looger LL, Porter LL. Inaccurate secondary structure predictions often indicate protein fold switching. Protein Sci. 2019;28(8):1487-1493.
- Mayrose I, Graur D, Ben-Tal N, Pupko T. Comparison of site- specific rate-inference methods for protein sequences: Empiri- cal bayesian methods are superior. Mol Biol Evol. 2004;21(9): 1781-1791.
- Seidel J, Hoffmann M, Ellis KE, et al. Maca is a second cyto- chrome c peroxidase of geobacter sulfurreducens. Biochemis- try. 2012;51(13):2747-2756.
- Chang YG, Cohen SE, Phong C, et al. Circadian rhythms. A protein fold switch joins the circadian oscillator to clock output in cyanobacteria. Science. 2015;349(6245):324-328.
- Kamisetty H, Ovchinnikov S, Baker D. Assessing the utility of coevolution-based residue-residue contact predictions in a sequence-and structure-rich era. Proc Natl Acad Sci U S A. 2013;110(39):15674-15679.
- Pinheiro F, Santos J, Ventura S. Alphafold and the amyloid landscape. J Mol Biol. 2021;433(20):167059.
- Strodel B. Energy landscapes of protein aggregation and con- formation switching in intrinsically disordered proteins. J Mol Biol. 2021;433(20):167182.
- Burmann BM, Knauer SH, Sevostyanova A, et al. An alpha helix to beta barrel domain switch transforms the transcription factor rfah into a translation factor. Cell. 2012;150(2):291-303.
- Giganti D, Albesa-Jove D, Urresti S, et al. Secondary structure reshuffling modulates glycosyltransferase function at the mem- brane. Nat Chem Biol. 2015;11(1):16-18.
- Mapelli M, Massimiliano L, Santaguida S, Musacchio A. The mad2 conformational dimer: Structure and implications for the spindle assembly checkpoint. Cell. 2007;131(4):730-743.
- Tseng R, Goularte NF, Chavan A, et al. Structural basis of the day-night transition in a bacterial circadian clock. Science. 2017;355(6330):1174-1180.
- Brown CJ, Johnson AK, Daughdrill GW. Comparing models of evolution for ordered and disordered proteins. Mol Biol Evol. 2010;27(3):609-621.
- Del Alamo D, Govaerts C, McHaourab HS. Alphafold2 predicts the inward-facing conformation of the multidrug transporter lmrp. Proteins. 2021;89(9):1226-1228.
- Kim AK, Porter LL. Functional and regulatory roles of fold- switching proteins. Structure. 2021;29(1):6-14.
- Li BP, Mao YT, Wang Z, et al. Clic1 promotes the progression of gastric cancer by regulating the mapk/akt pathways. Cell Physiol Biochem. 2018;46(3):907-924.
- Milton RH, Abeti R, Averaimo S, et al. Clic1 function is required for beta-amyloid-induced generation of reactive oxy- gen species by microglia. J Neurosci. 2008;28(45):11488-11499.
- Chen N, Das M, LiWang A, Wang LP. Sequence-based predic- tion of metamorphic behavior in proteins. Biophys J. 2020; 119(7):1380-1390.
- Kim AK, Looger LL, Porter LL. A high-throughput predictive method for sequence-similar fold switchers. Biopolymers. 2021; 112(10):e23416.
- Cai M, Huang Y, Shen Y, et al. Probing transient excited states of the bacterial cell division regulator mine by relaxation dis- persion nmr spectroscopy. Proc Natl Acad Sci U S A. 2019; 116(51):25446-25455.
- Dishman AF, Tyler RC, Fox JC, et al. Evolution of fold switching in a metamorphic protein. Science. 2021;371(6524): 86-90.
Cited by
Sorted by citations
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
Scalable emulation of protein equilibrium ensembles with generative deep learning
Lewis, S., Hempel, T., Jiménez-Luna, J., Gastegger, M., Xie, Y., Foong, A. Y. K., Satorras, V. G., Abdin, O., Veeling, B. S., Zaporozhets, I., et al. (2025) · Science · doi:10.1126/science.adv9817 · Cited by 335 · see citing works
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
The power and pitfalls of AlphaFold2 for structure prediction beyond rigid globular proteins
Agarwal, V., & McShan, A. C. (2024) · Nature Chemical Biology · doi:10.1038/s41589-024-01638-w · Cited by 155 · see citing works
“These deviations can be broadly categorized into cases with (1) inaccurate secondary, tertiary and/or quaternary the accuracy of the standard implementation of AF2 against different types of nonglobular molecular targets, which could limit its potential applications 15,16 .”
BackgroundProtein structure prediction has reached the single-structure frontier
Lane, T. J. (2023) · Nature Methods · doi:10.1038/s41592-022-01760-4 · Cited by 125 · see citing works
Zhang, 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 · see citing works
“Another noted limitation of AF2 is that it is designed to predict a static conformation given an input sequence 22,23 .”
Ohnuki, J., Jaunet-Lahary, T., Yamashita, A., & Okazaki, K.-I. (2024) · The Journal of Physical Chemistry Letters · doi:10.1021/acs.jpclett.3c03052 · Cited by 14 · see citing works
Severe deviation in protein fold prediction by advanced AI: a case study
López-Sagaseta, J., & Urdiciain, A. (2025) · Scientific Reports · doi:10.1038/s41598-025-89516-w · Cited by 10 · see citing works
Lidbrink, S. E., Howard, R. J., Haloi, N., & Lindahl, E. (2025) · PLOS Computational Biology · doi:10.1371/journal.pcbi.1013187 · Cited by 9 · see citing works
A large‐scale evolutionary and structural analysis of <scp>CLC</scp> channels and transporters
Mishra, A., Díaz Vázquez, G., & Robertson, J. L. (2025) · Protein Science · doi:10.1002/pro.70389 · Cited by 3 · see citing works
“Indeed, AF2 has been shown to be less effective at predicting the structure of fold-switching proteins (Chakravarty & Porter, 2022), revealing its inherent bias toward structures in the training set.”
Scalable emulation of protein equilibrium ensembles with generative deep learning
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., et al. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.12.05.626885 · Cited by 349 · see citing works
- Authors
- Chakravarty, D., & Porter, L. L.
- Affiliations
- National Institutes of Health (US)
- Corresponding author
- Porter, Lauren L.
- Venue
- Protein Science
- Publisher
- Wiley
- Published
- 2022-05-26
- Volume
- 31
- Issue
- 6
- Pages
- e4353-e4353
- ISSN
- 0961-8368 (ISSN-L)
- 1469-896X
- PMC
- PMC9134877
- OpenAlex
- W4281551093
- Cited by
- 193
- References
- 9
- Citations by year
- 2026: 35 · 2025: 51 · 2024: 58 · 2023: 37 · 2022: 12
- FWCI
- 12.90
- Type
- journal article
- Language
- en
- Open access
- bronze
- doi.org
- Locations
- Protein Science (publishedVersion)
- bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion)
- PubMed Central (publishedVersion, public-domain)
- PubMed (publishedVersion)
- Funders
- National Institutes of Health
- U.S. National Library of Medicine
- 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, protein structure prediction, structural heterogeneity, intrinsically disordered regions, prediction confidence, sequence conservation, multiple sequence alignment, protein conformational ensembles
- MeSH terms
- Sequence Alignment, Intrinsically Disordered 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
