Using AlphaFold to predict the impact of single mutations on protein stability and function
Open accessPublished version of a preprint
The DOI matches its Crossref record; 33 of the 33 references it cites were found in a registry or an index; 3 of the 9 sources its claims cite were checked for support.
Abstract
AlphaFold changed the field of structural biology by achieving three-dimensional (3D) structure prediction from protein sequence at experimental quality. The astounding success even led to claims that the protein folding problem is “solved”. However, protein folding problem is more than just structure prediction from sequence. Presently, it is unknown if the AlphaFold-triggered revolution could help to solve other problems related to protein folding. Here we assay the ability of AlphaFold to predict the impact of single mutations on protein stability (ΔΔG) and function. To study the question we extracted the pLDDT and metrics from AlphaFold predictions before and after single mutation in a protein and correlated the predicted change with the experimentally known ΔΔG values. Additionally, we correlated the same AlphaFold pLDDT metrics with the impact of a single mutation on structure using a large scale dataset of single mutations in GFP with the experimentally assayed levels of fluorescence. We found a very weak or no correlation between AlphaFold output metrics and change of protein stability or fluorescence. Our results imply that AlphaFold may not be immediately applied to other problems or applications in protein folding.
Overview
CoChat Summary
TL;DR
The paper tests whether AlphaFold confidence metrics can predict the effect of single mutations on protein stability (ΔΔG) and GFP fluorescence, using experimental mutation datasets. It finds weak or no useful correlation, concluding AlphaFold outputs are not directly reliable for these prediction tasks.
Key findings
- For 1154 mutations in 73 proteins, the correlation between mutation-induced change in local confidence and experimental stability change was weak: PCC = -0.17 ± 0.03 for ΔpLDDT, while Δ<pLDDT> was 0.02 ± 0.03.
- A multiple linear regression combining ΔpLDDT and Δ<pLDDT> did not improve prediction much: 0.12 ± 0.05 on the training set and 0.20 ± 0.04 on the test set.
- For 796 GFP single mutants, AlphaFold metrics also showed only weak association with function: PCC = 0.17 ± 0.03 for ΔpLDDT and 0.16 ± 0.04 for Δ<pLDDT>.
- Among destabilizing mutations, 87% had negative ΔpLDDT, but this still did not produce a meaningful correlation with ΔΔG.
Methods
- Experimental stability data came from ThermoMutDB v1.3; after filtering for single-point mutations measured at pH 3–9, 293–300 K, and proteins <250 aa, the authors analyzed 1154 mutations in 73 proteins.
- GFP function data came from a published dataset; the authors randomly selected 796 single mutants and used fluorescence as the functional readout.
- Wild-type and mutant structures were obtained from the AlphaFold Protein Structure Database or predicted with standalone AlphaFold, and analysis focused on ΔpLDDT and Δ<pLDDT>; the stability dataset was split into training/test sets by a 50% sequence identity threshold and analyzed with multiple linear regression.
Limitations
- The paper states that AlphaFold was not designed or validated to predict mutation effects on stability or function, so direct use is not expected to work well.
- The authors note that the available ΔΔG data are limited in size (they describe experimentally measured ΔΔG values as being on the order of 10,000), which they argue constrains deep-learning prediction; this is an interpretation made in the paper.
- The evidence is limited to the datasets studied here: single-point mutations in ThermoMutDB proteins and GFP mutants, so broader generalization is not stated in the paper.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures


Verification
Claims the paper makes
3 Valid · 0 Not valid · 3 Not checked · 19 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Thus, using AlphaFold models instead of PDB structures does not make ΔΔG predictions more accurate [15]”
§ Using AlphaFold to predict the impact of single mutations on protein stability and function > Discussion
What the sources say
- Partly supportedcites [15] Pak MA, Ivankov DN. Best templates outperform homology models in predicting the impact of mutations on protein stability. Bioinformatics 2022 38(18): 4312– 4320. doi: 10.1093/bioinformatics/btac515 35894930 sourceExperimental — accuracy not yet measured
Section 3.6 (Protein stability change prediction with AlphaFold models) states that 'utilization of predicted protein structure does not improve the accuracy of protein stability change prediction by FoldX', and the Discussion ('Surprisingly, the conclusion holds true even for the models made by AlphaFold') supports the direction of the claim. However, the support is bounded in ways the citing sentence drops: it applies only to FoldX as the ΔΔG predictor, and rests on just two CASP14 AlphaFold models with 1200 random mutations, where correlation coefficients were statistically indistinguishable but prediction error was actually greater for the AlphaFold models (Figure 5); the authors explicitly caution that 'this conclusion is compromised by the limited amount of structures available for evaluation'. So the generalized, unhedged claim 'using AlphaFold models instead of PDB structures does not make ΔΔG predictions more accurate' overstates the scope and certainty of what this source shows. (LLM, conf=0.75)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The correlation coefficient is -0.17 ± 0.03 (p-value = 10-8) for ΔpLDDT and 0.02 ± 0.03 (p-value = 0.44) for the Δ<pLDDT>.”
§ Using AlphaFold to predict the impact of single mutations on protein stability and function > Results > Correlation between ΔΔG and ΔpLDDT values
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Expectedly, the destabilizing effect of mutations was associated with decreasing pLDDTs: 87% of destabilizing mutations had negative ΔpLDDT (p-value = 10−22).”
§ Using AlphaFold to predict the impact of single mutations on protein stability and function > Results > Relationship of ΔpLDDT and amino acid properties
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“However, there was no correlation between ΔpLDDT and ΔΔG for that 87% of mutations with negative ΔpLDDT.”
§ Using AlphaFold to predict the impact of single mutations on protein stability and function > Results > Relationship of ΔpLDDT and amino acid properties
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We found a very weak or no correlation between AlphaFold output metrics and change of protein stability or fluorescence.”
- Also stated in: § Correlation between ΔΔG and ΔpLDDT values · plain
“We did not observe a pronounced correlation between the mutation effect and the difference in confidence metrics (Fig 1).”
- Also stated in: § Discussion · plain
“We found a weak correlation of -0.17 ± 0.03 between ΔpLDDT and ΔΔG associated with specific mutations.”
§ Using AlphaFold to predict the impact of single mutations on protein stability and function > Abstract
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § Correlation between ΔΔG and ΔpLDDT values · plain
14 more claims
Create a free account to see all 25 claims and to verify the 1 reference still open.
- Probably not a substantial claim. 2
- It may state two separate claims as one. 4
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 source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“studies have reported an inability of AF2 to predict the effects of mutations on protein structure and stability14,49,50”
From The power and pitfalls of AlphaFold2 for structure prediction beyond rigid globular proteins (2024) · cites this paper as [50]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“Beyond predicting the predominant state of naturally occurring proteins, we wanted to test the ability of AF-Cluster to predict effects of point mutations, a task that AF2 in its default settings has not achieved40.”
From Predicting multiple conformations via sequence clustering and AlphaFold2 (2024) · cites this paper as [40]
Citations
Cites
Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing
Benjamini, Y., & Hochberg, Y. (1995) · Journal of the Royal Statistical Society Series B: Statistical Methodology · doi:10.1111/j.2517-6161.1995.tb02031.x · Cited by 109,482 · cited 1×
“All correlations were adjusted for multiple hypotheses testing by Benjamini-Hochberg correction [22].”
MethodBasic local alignment search tool
Altschul, S. F., Gish, W., Miller, W., Myers, E. W., & Lipman, D. J. (1990) · Journal of Molecular Biology · doi:10.1006/jmbi.1990.9999 · Cited by 98,669 · cited 4×
“To identify the sequence identities between the proteins in the dataset of mutations we performed protein BLAST [18] search of protein sequences against themselves.”
MethodInfluentialHighly 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 3×
“AlphaFold is widely claimed to have revolutized protein 3D structure prediction from protein sequence, a 50-years long-standing challenge of protein physics and structural bioinformatics [1].”
BackgroundMethodInfluentialBerman, H. M. (2000) · Nucleic Acids Research · doi:10.1093/nar/28.1.235 · Cited by 40,196
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 2×
“Next, AlphaFold in collaboration with EMBL-EBI constructed the structure models for the whole protein sequence space [8].”
MethodKnowledge‐based protein secondary structure assignment
Frishman, D., & Argos, P. (1995) · Proteins: Structure, Function, and Bioinformatics · doi:10.1002/prot.340230412 · Cited by 2,523 · cited 1×
“Information on solvent accessibility was taken from Stride [19].”
BackgroundGuerois, R., Nielsen, J. E., & Serrano, L. (2002) · Journal of Molecular Biology · doi:10.1016/s0022-2836(02)00442-4 · Cited by 1,728 · cited 3×
“Similarly, the same AlphaFold metrics had a very weak correlation with the impact of single mutations on protein function, fluorescence, of GFP. Recent results show that the use of AlphaFold models instead of template structures does not improve ΔΔG prediction by FoldX [12] [13].”
ResultInfluentialHarnessing protein folding neural networks for peptide–protein docking
Tsaban, T., Varga, J. K., Avraham, O., Ben-Aharon, Z., Khramushin, A., & Schueler-Furman, O. (2022) · Nature Communications · doi:10.1038/s41467-021-27838-9 · Cited by 1,205
Harnessing protein folding neural networks for peptide-protein docking
(2021) · doi:10.21203/rs.3.rs-781411/v1 · Cited by 1,203
Mariani, V., Biasini, M., Barbato, A., & Schwede, T. (2013) · Bioinformatics · doi:10.1093/bioinformatics/btt473 · Cited by 1,138 · cited 1×
“LDDT ranges from 0 to 100 and is a superposition-free metric indicating to what extent the protein model reproduces the reference structure [23].”
Background
Cited by
Items 201–300 of 320
Progresses in genetic testing in congenital heart disease
Gao, H., Liu, Y., Sheng, W., Shou, W., & Huang, G. (2024) · Medicine Plus · doi:10.1016/j.medp.2024.100028 · Cited by 4
The origin of mutational epistasis
Vila, J. A. (2024) · European Biophysics Journal · doi:10.1007/s00249-024-01725-9 · Cited by 6
van Aalst, E. J., & Wylie, B. J. (2024) · Protein Science · doi:10.1002/pro.70013 · Cited by 5
“Previous reports found no correlation between pLDDT and experimental ΔΔ G or protein function, but the main thrust of these works was ultimately an attempt to find correlations to address protein folding (Buel & Walters, 2022; Pak et al., 2023).”
InfluentialNapiorkowska, M., Fischer, K., Penner, M., Knyphausen, P., & Hollfelder, F. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.11.20.567881 · Cited by 3
“…the limitations of currently available methods in predicting the effects of single point mutations on binding (as recently pointed out by others 73,74 ) and emphasizes the need for improved high-711 throughput methods in experimental protein engineering to generate large datasets that can fuel…”
Swapna, G. V. T., Dube, N., Roth, M. J., & Montelione, G. T. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.07.15.603529 · Cited by 3
Recent advances in the integration of protein mechanics and machine learning
Chen, Y., & Chang, S. (2024) · Extreme Mechanics Letters · doi:10.1016/j.eml.2024.102236 · Cited by 3
Seman‐Kamarulzaman, A., Muhammad, N. A. N., Othman, N. W., Jalinas, J., Hassan, N. I., Rahmat, N. L., & Hassan, M. (2025) · Archives of Insect Biochemistry and Physiology · doi:10.1002/arch.70039 · Cited by 2
Dummunee, K., Parry, R. H., Redecke, L., Varjak, M., Brennan, B., Kohl, A., & McFarlane, M. (2025) · Journal of Biological Chemistry · doi:10.1016/j.jbc.2025.108332 · Cited by 5
Radler, C. N., Wang, T., LeGate, J. F., Crone, L., Deo, P., Wortley, J., Moore, P., Bryant, G., Smitherman, K., & Sathyamoorthy, M. (2025) · International Journal of Molecular Sciences · doi:10.3390/ijms26136535 · Cited by 2
“However, the use of AlphaFold and other computational methods to determine the effect of mutations is controversial, with recent studies highlighting difficulties in obtaining accurate representations of the stability and function of mutated proteins [76–79].”
A Survey of Current Status in AI-Based Topology Prediction of Transmembrane Proteins
Bathla, D., Mishra, R., & Ahmad, S. (2025) · Methods in molecular biology · doi:10.1007/978-1-0716-4662-5_6 · Cited by 3
Predicted peptide scaffolds for drug screening in endometrial cancer organoids
Zhang, M., Wan, Y., & Li, D. (2025) · Scientific Reports · doi:10.1038/s41598-025-21282-1 · Cited by 3
Sebastiano, M. R., Apprato, G., Francisco, S., Ermondi, G., & Caron, G. (2026) · Medicinal Research Reviews · doi:10.1002/med.70038 · Cited by 5
Zeuzem, N., Quilan, M., Valentin, M. D., Baert‐Desurmont, S., Horlacher, M., Valle, A. D., Esperón, P., Neffa, F., Machado-Lopes, T. M. B., Nascimento, I., et al. (2026) · European Journal of Human Genetics · doi:10.1038/s41431-026-02072-3 · Cited by 2
Durumeric, A. E. P., McCarty, S., Smith, J., Köhler, J., Elez, K., Raich, L., Suriana, P., & Sztain, T. (2026) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.6c00332 · Cited by 2
“60,61 Furthermore, many mutations prevent successful folding, presenting a challenge for structure prediction.”
Standardized numbering and alignment of the KPC family of β-lactamases
Brunetti, F., Marshall, S. H., Bethel, C. R., Hujer, A. M., Hujer, K. M., Haider, S. M., Bush, K., Bradford, P. A., Kreiswirth, B., SPENCER, J. M., et al. (2026) · Antimicrobial Agents and Chemotherapy · doi:10.1128/aac.01868-25 · Cited by 5
GTExome: Modeling commonly expressed missense mutations in the human genome
Hoffman, J., Tan, H., Sandoval-Cooper, C., de Villiers, K., & Reed, S. M. (2024) · PLoS ONE · doi:10.1371/journal.pone.0303604 · Cited by 2
“AlphaFold is a deep learning method that predicts 3D protein structure from the amino acid sequence based on known protein structures with similar sequences [6].”
MethodZhang, Z., Zhang, H., & Dai, W. J. (2025) · Journal of Pharmacology and Experimental Therapeutics · doi:10.1016/j.jpet.2025.103402 · Cited by 2
TRain: T-cell receptor automated immunoinformatics
Seamann, A., Bennett-Boehm, M. M. C., Ehrlich, R., Gil, A., Selin, L. K., & Ghersi, D. (2025) · BMC Bioinformatics · doi:10.1186/s12859-025-06074-8 · Cited by 4
“However, it is important to note that predictive performance of AlphaFold2 on discriminating the effect of point mutations has been shown to be poor [19].”
Vo, D. N. K., Ho, H. P. T., Wu, L., Chen, Y., Trinh, H. T. V., Lin, T., Lim, Y., Tsai, K., & Tsai, M. (2025) · International Journal of Biological Macromolecules · doi:10.1016/j.ijbiomac.2025.142179 · Cited by 2
Tevenvirinae phages encode a family of LSm-like fold proteins
Kim, J., Tai, C., Livingston, N. K., Patterson-West, J., Stearns, O., Son, B., & Hinton, D. M. (2025) · Computational and Structural Biotechnology Journal · doi:10.1016/j.csbj.2025.05.028 · Cited by 2
Gao, W., Bai, R., & Ling, S. (2025) · Trends in Food Science & Technology · doi:10.1016/j.tifs.2025.105193 · Cited by 7
Myronidi, I., Ring, A., Wú, F., & Ljungdahl, P. O. (2023) · The Journal of Cell Biology · doi:10.1083/jcb.202208060 · Cited by 2
Szulc, N., Gąsior‐Głogowska, M., Żyłka, P., Szefczyk, M., Wojciechowski, J. W., Żak, A. M., Dyrka, W., Kaczorowska, A., Burdukiewicz, M., Tarek, M., et al. (2024) · Spectrochimica Acta Part A Molecular and Biomolecular Spectroscopy · doi:10.1016/j.saa.2024.124094 · Cited by 3
Mahmud, S., Morehead, A., & Cheng, J. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.10.03.560758 · Cited by 4
“That is, given the sequence of a wild-type protein or its mutant with a single amino acid sub-stitution, AlphaFold still cannot accurately predict their structural difference [6, 7].”
Cerebra: A computationally efficient framework for accurate protein structure prediction
Hu, J., Wang, W., Zhang, B., Cao, Y., & Gong, H. (2026) · Protein Science · doi:10.1101/2024.02.02.578551 · Cited by 6
An integrative approach to protein sequence design through multiobjective optimizationPreprint
Hong, L., & Kortemme, T. (2024) · bioRxiv · doi:10.1101/2024.03.01.582670 · Cited by 20
“These results suggest that a naive application of a genetic algorithm is likely to be outperformed by a single-pass inverse folding model, and AlphaFold2 is, on its own, a poor mutational effect predictor [58,61].”
Dominik, N., Efthymiou, S., Record, C. J., et al. (2024) · medRxiv · doi:10.1101/2024.05.10.24306768 · Cited by 2
Galeazzi, F. M., da Silva, G. M., Arantes, P. R., Varghese, I., Shukla, A., & Rubenstein, B. M. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.05.09.653048 · Cited by 3
AlphaFold reveals but sometimes distorts an organizational principle of protein foldingPreprint
Clore, M. F., Thole, J. F., Dontha, S., Sharma, P., Greenberg, N., Strub, M., Starich, M. R., Ott, C. M., Jensen, D. R., Volkman, B. F., et al. (2026) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2025.12.30.697132 · Cited by 5
“While remarkably successful at predicting single structures and oligomeric assemblies from amino acid sequences ( 11-13 ), AlphaFold remains unreliable at selecting among alternative conformational states ( 8, 14-17 ).”
Feldman, J. D., Brogi, M., & Skolnick, J. (2026) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2026.02.25.708002 · Cited by 2
Long, E., & Monroe, G. (2026) · BMC Genomics · doi:10.1186/s12864-026-12674-2 · Cited by 3
“It is still yet unknown how well these in silico reconstruction of protein structures can be used to estimate effects of single point mutations [4], however, some studies have shown high correlations between AlphaFold predicted importance and disease pathogenicity of non-synonymous mutations [5].”
(2025) · doi:10.1371/journal.pcbi.1013216 · Cited by 4
“…caveat is AlphaFold2 [23], a black box model that achieved near-experimental accuracy on the CASP14 protein folding challenge [24] but is unable to predict the impact of structure-disrupting mutations, which are frequently associated with protein aggregation, misfolding, and dysfunction [25, 26].”
Building protein structure-specific rotamer libraries
Grybauskas, A., & Gražulis, S. (2023) · Bioinformatics · doi:10.1093/bioinformatics/btad429 · Cited by 3
“For example, it was noticed with AlphaFold2 when trying to predict the impact of the point mutations to the protein stability, the energy values did not correlate with other widely used methods (Pak et al. 2023).”
ResultSam-Sam Association Between EphA2 and SASH1: In Silico Studies of Cancer-Linked Mutations
Vincenzi, M., Mercurio, F. A., Autiero, I., & Leone, M. (2025) · Molecules · doi:10.3390/molecules30030718 · Cited by 2
Next-generation nephrology: part 1—an aid for genetic and genomic testing in pediatric nephrology
Gupta, A., Jayasinghe, K., Majmundar, A. J., Mann, N., Sinha, R., Sampson, M. G., & Quinlan, C. (2025) · Pediatric Nephrology · doi:10.1007/s00467-025-06697-2 · Cited by 3
A Brief Introduction to Pharmacogenomics and Personalized Medicine in the Drug Design Context
Ferreira, G. M., Hirata, M. H., Cappello, T. P., Dagli-Hernandez, C., & Fukushima, A. R. (2024) · Computer-aided drug discovery and design · doi:10.1007/978-3-031-76718-0_3 · Cited by 1
Yashinov, A., Zou, X., Hang, J., Liu, Z., Song, F., Zeng, Y., Yang, Y., Xia, F., Tang, F., Shi, W., et al. (2025) · Chemical Research in Chinese Universities · doi:10.1007/s40242-025-5013-0 · Cited by 1
Walla, B., Dietrich, A.-M., Brames, E., Bischoff, D., Fritzsche, S., Castiglione, K., Janowski, R., Niessing, D., & Weuster‐Botz, D. (2025) · Bioengineering · doi:10.3390/bioengineering12060561 · Cited by 2
Enzyme sequence optimisation via Gromov-Wasserstein Autoencoders integrating MSA techniques
Wang, X., Li, Y., Hou, X., & Liu, H. (2025) · Journal of Enzyme Inhibition and Medicinal Chemistry · doi:10.1080/14756366.2025.2524742 · Cited by 1
Li, Y., Xu, S., Zhang, M., Yang, X., & Wei, Z. (2025) · International Journal of Molecular Sciences · doi:10.3390/ijms26136451 · Cited by 3
“Wild-type protein models were obtained from the AlphaFold database using UniProt “https://www.uniprot.org/ (accessed on 15 March 2025)” access codes [56], and interaction schemes were predicted using the AlphaFold Server.”
AFToolkit: a framework for molecular modeling of proteins with AlphaFold-derived representations
Sindeeva, M., Telepov, A., Ivanisenko, N. V., Shashkova, T., Khrabrov, K., Tsypin, A., Kadurin, A., & Kardymon, O. L. (2025) · Briefings in Bioinformatics · doi:10.1093/bib/bbaf324 · Cited by 1
“…and template features is one of the reasons for AlphaFold2’s limited predictive power in evaluating the effects of amino acid point mutations [36], possibly due to the predicted structures for mutants being highly correlated with wildtype structures and therefore not accounting for the…”
Afshar, Y., Naseri, A., Amoozadehsamakoosh, A., Naseri, A., Abrishami, M., Yazdanpanah, N., Saleki, K., & Rezaei, N. (2025) · Journal of Biomolecular Structure and Dynamics · doi:10.1080/07391102.2025.2534020 · Cited by 1
“Taken together, findings showed a poor or a lack of association between AlphaFold output metrics and alterations in protein stability, suggesting that AlphaFold may not currently be applied to other issues or usages in protein folding (Pak et al., 2023).”
AlphaFold 3 accurately models natural variants of Helicobacter pylori catalase KatA
Gómez, A. A., & Baylink, A. (2025) · Microbiology Spectrum · doi:10.1128/spectrum.00670-25 · Cited by 6
Gaertner, N. F., Maio, F., Arroyo-Mateos, M., Luna, A. P., Sabarit, B., Kwaaitaal, M., Eltschkner, S., Prins, M., Bejarano, E. R., & van den Burg, H. A. (2025) · Journal of Virology · doi:10.1128/jvi.01286-25 · Cited by 1
Protein Engineering and Drug Discovery: Importance, Methodologies, Challenges, and Prospects
Mohammed, A., Ibrahim, N. A., & Basher, N. S. (2025) · Biomolecules · doi:10.3390/biom15111628 · Cited by 3
“However, AlphaFold’s limitations stem from its training on stable proteins, which can lead to predictions that may not account for potential instability [182,183].”
Chen, W., Hou, D., Liu, Y., & Zhang, X. H. (2025) · Communications Biology · doi:10.1038/s42003-025-09206-1 · Cited by 3
“Thoughsingleorlimitedmutationsusually willbesmoothedoutduetotheAlphaFoldalgorithm 22 , our mutations of C1669A and C1670A included a signi fi cant structural change (the removal ofadisul fi debond)whichislikelytoshifttheAlphaFold3predictingresults, offering different predictions between WT A2 and…”
Schwabe, J., Feria, J. M., Glatter, T., Blumenkamp, P., Rupp, O., Goesmann, A., Valvano, M. A., & Søgaard‐Andersen, L. (2025) · mBio · doi:10.1128/mbio.02986-25 · Cited by 2
“The predicted interface template modeling score (ipTM) of 0.79 is close to the stringent 0.80 cutoff (80), supporting an overall accurate prediction of the dimer interface (Fig.”
InfluentialMirabal, D. M. R., Jiao, S., Mengel, S. D., Segalman, R. A., & Shell, M. S. (2025) · The Journal of Chemical Physics · doi:10.1063/5.0299938 · Cited by 1
Repetitive proteins that undergo large conformational changes evade structural prediction algorithms
Chang, M. P., Jin, T., Gudinas, A. P., Fernández, D., Alexander‐Katz, A., Matsui, T., & Mai, D. J. (2025) · The Journal of Chemical Physics · doi:10.1063/5.0304777 · Cited by 2
Homology Modeling in Protein Structure Predictions
Thai, K., Nguyen, T.-K., Nguyen, D.-N., & Le, M.-T. (2025) · Springer handbooks · doi:10.1007/978-3-031-81728-1_26 · Cited by 2
Dematophora necatrix: From taxonomy to molecular advances
Carstens, M., Pliego, C., Norval, A., & van den Berg, N. (2026) · Fungal Biology Reviews · doi:10.1016/j.fbr.2026.100471 · Cited by 1
Bridging structure and function: artificial intelligence-based modelling of kidney proteins
Wu, S., Wang, W., Zhou, Z. H., Scalzo, F., & Kurtz, I. (2026) · Nature Reviews Nephrology · doi:10.1038/s41581-026-01060-6 · Cited by 1
“These limitations are well documented, and show that repurposing AF2 for stability prediction performs poorly 106,108 .”
Beyond accuracy: comprehensive alignment for AI-driven molecular design
Zhang, Q. H., Zhuang, X., Zhou, C., Zhu, Y., Xu, T., Li, T., Liu, D., Liu, S., Ding, K., Li, M., et al. (2026) · AI for Science · doi:10.1088/3050-287x/ae5f37 · Cited by 4
AI-driven insights into protein misfolding and innate immunity in neurodegenerative diseases
Deng, H. X., Cao, J. L., Yao, W., Jiang, S. J., Fang, Q. Q., Zhu, B. Y., & Jiang, Y. J. (2026) · Frontiers in Immunology · doi:10.3389/fimmu.2026.1814357 · Cited by 2
Analysing open-source protein folding models for nanobody binding prediction
Vogt, Y., Roßberg, R., Habermann, J., Kluge, P., Xu, W., Naouar, M., Hirschberg, S., Miething, C., Bödecker, J., Kalweit, M., et al. (2026) · Frontiers in Bioinformatics · doi:10.3389/fbinf.2026.1858891 · Cited by 1
“…2021) and its successor AF3 (Abramson et al., 2024) have been shown to poorly capture the effects of missense mutations (Buel and Walters, 2022; Pak et al., 2021), and to struggle with antibody–antigen complex prediction (Abramson et al., 2024), as well as nanobody docking (Hitawala and Gray,…”
Emergence Biases in Molecular Evolution
Fuqua, T., & Vakirlis, N. (2026) · Genome Biology and Evolution · doi:10.1093/gbe/evag195 · Cited by 2
Structural and functional consequences of IL33 and IL1RL1 coding variants on IL-33/ST2 signaling
Gelon, L., Roga, S., Nahoum, V., Pasquet, M., Prade, N., Dufrechou, S., Largeaud, L., Delabesse, É., Voisin, S., Girard, J., et al. (2026) · Genes and Immunity · doi:10.1038/s41435-026-00410-5 · Cited by 2
Ma, M., Hou, I., Gao, J., Li, C., Bai, Y., & Liu, X. (2022) · IEEE International Conference on Bioinformatics and Biomedicine (BIBM) · doi:10.1109/bibm55620.2022.9995589 · Cited by 1
“There are some studies investigating AlphaFold2’s ability in predicting the effects of genetic variants [5, 6].”
BackgroundDesign of stable circular permutants of the GroEL chaperone apical domain
Melnik, T. N., Majorina, M. A., Vorobeva, D. E., Nagibina, G. S., Veselova, V. R., Glukhova, K. A., Pak, M. A., Ivankov, D. N., Uversky, V. N., & Melnik, B. S. (2024) · Cell Communication and Signaling · doi:10.1186/s12964-023-01426-4 · Cited by 1
“This observation harmonizes with previous findings concerning single-point mutations [49, 52].”
ResultSelf-citationGálvez-Ramírez, A., González‐Valdez, A., Hernández‐Ochoa, B., Canseco‐Ávila, L. M., López-Roblero, A., Arreguín‐Espinosa, R., de la Cruz, V. P., Hernández-Urzúa, E., Cárdenas‐Rodríguez, N., Enríquez‐Flores, S., et al. (2024) · International Journal of Molecular Sciences · doi:10.3390/ijms252312556 · Cited by 2
“For example, Pak et al. [36] explored the capacity for AF2 predictions to evaluate the impacts of mutations on protein stability changes ( ∆∆ G) despite the software having a disclaimer stating that the routine “ has not been validated for predicting the effect of mutations ”…”
Protein‐Variant‐Phenotype Study of NBAS Using AlphaFold in the Aspect of SOPH Syndrome
Жожиков, Л., Vasilyev, F. F., & Максимова, Н. Р. (2024) · Proteins Structure Function and Bioinformatics · doi:10.1002/prot.26764 · Cited by 1
High‐Throughput Evaluation of Natural Diversity of F‐Type ATP Synthase Rotor Ring Stoichiometries
Osipov, S. D., Zinovev, E. V., Anuchina, A. A., Kuzmin, A., Minaeva, A. V., Ryzhykau, Y. L., Vlasov, A. V., & Gushchin, I. (2025) · Proteins Structure Function and Bioinformatics · doi:10.1002/prot.26790 · Cited by 3
Savvateeva, L. V., Chepikova, O. E., Solonkina, A. D., Sakharov, A. A., Gorokhovets, N. V., Golovin, A. V., & Zamyatnin, A. A. (2025) · Pharmaceuticals · doi:10.3390/ph18040592 · Cited by 2
“It has been shown previously that for point contacts, there is no correlation between prediction quality and the experimental relative change in Gibbs energy with substrate binding values [36].”
Walker, D. R., Fujimura, G., Vanegas, J. M., & Barbar, E. J. (2025) · Frontiers in Molecular Biosciences · doi:10.3389/fmolb.2025.1531793 · Cited by 1
Structural Variability of Pfam Domains Based on Alphafold2 Predictions
Poretsky, E., Andorf, C. M., & Sen, T. Z. (2025) · Proteins Structure Function and Bioinformatics · doi:10.1002/prot.70021 · Cited by 2
Prescott, L. (2022) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2022.01.22.477317 · Cited by 1
The impact of genetically controlled splicing on exon inclusion and protein structurePreprint
Einson, J. E., Minaeva, M., Rafi, F., & Lappalainen, T. (2022) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2022.12.05.518915 · Cited by 1
Corre, M.-H., Rey, B., David, S. C., et al. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.08.30.555488 · Cited by 1
What makes the effect of protein mutations difficult to predict?Preprint
Redestig, H., van der Flier, F. J., Estell, D., Pricelius, S., Dankmeyer, L., Thans, S. V. S., Mulder, H., Otsuka, R., Goedegebuur, F., Lammerts, L., et al. (2023) · Research Square · doi:10.21203/rs.3.rs-3415956/v1 · Cited by 1
Lai, H.-S., & Chen, C. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.11.25.568523 · Cited by 1
“Incontemporary research, the predominant focus has been predicting the pathogenicity or thermodynamic free energy of nsSNVs rather than their structural changes (Pak et al., 2023).”
BackgroundCerebra: A Computationally Efficient Framework for Accurate Protein Structure PredictionPreprint
Gong, H., Hu, J., & Wang, W. (2024) · Research Square · doi:10.21203/rs.3.rs-3995729/v1 · Cited by 1
AlphaFold 3 accurately models natural variants of Helicobacter pylori catalase KatAPreprint
Perkins, A. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.06.02.657526 · Cited by 1
“A recent study showed that AlphaFold’s predictions for the energetic consequences of single mutations poorly correlate with experimentally determined protein stabilities 9,20 .”
An improved model for prediction of de novo designed proteins with diverse geometriesPreprint
Orr, B., Crilly, S. E., Akpinaroglu, D., Zhu, E., Keiser, M. J., & Kortemme, T. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.06.02.657515 · Cited by 2
“For natural proteins, it has been shown previously that AF2 is not sensitive to point mutations 30 .”
Discapacidad y comunicación: propuesta en desarrollo. Informe
Gallardo, E. M., Rojas, N., Matos, H. M., & de Pietrosemoli, L. (2008) · Lengua y habla · Cited by 2
Mutafy: A webserver to identify high quality mutant protein structures in the Protein Data BankPreprint
Ness, D., Hu, J., Kalia, M., Dobson, R., Al‐Chalabi, A., & Iacoangeli, A. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.03.22.533870
“However, in-silico generated mutations do not account for the effect of mutations on folding and might not accurately reflect the conformational and functional changes caused by the mutations (8,9).”
BackgroundLi, C., Hou, I., Ma, M., Wang, G., Bai, Y., & Liu, X. (2023) · Frontiers in Bioinformatics · doi:10.3389/fbinf.2023.1122559
“Recent studies have tried to examine the ability of these tools individually to evaluate the impact of single amino-acid variants (SAVs), but reported con fl icting results (Pak et al., 2021; Caswell et al., 2022).”
BackgroundMunir, A., Akram, A. M., Jamil, K., & Tahir, A. (2023) · Pakistan BioMedical Journal · doi:10.54393/pbmj.v6i05.881 · Cited by 1
Gracia, B., Montes, P., Gutierrez‐Barrera, A. M., Arun, B., & Karras, G. I. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.09.14.557795
“However, these structure predictions breakdown when considering the impact of mutation on protein structure and function, likely because structure predictions are based on well-behaved “wild-type” proteins 25 .”
Background338 Features of genotype-phenotype correlation in siblings with cystic fibrosis
Simonov, M., Krasnovidova, A. E., Simonova, O. I., Gorinova, Y. V., Петровна, Ч. В., Savostyanov, K. V., & Pushkov, A. A. (2023) · Journal of Cystic Fibrosis · doi:10.1016/s1569-1993(23)01265-1
Deciphering GB1’s Single Mutational Landscape: Insights from MuMi Analysis
Guclu, T. F., Atilgan, A. R., & Atilgan, C. (2024) · The Journal of Physical Chemistry B · doi:10.1021/acs.jpcb.4c04916
Li, D., Zhu, Y., Zhang, W., Liu, J., Yang, X., Liu, Z., & Wei, D. (2024) · Interdisciplinary Sciences Computational Life Sciences · doi:10.1007/s12539-024-00662-7 · Cited by 1
AlphaFold2, SPINE-X, and Seder on Four Hard CASP Targets
Faraggi, E., Jernigan, R. L., & Kloczkowski, A. (2024) · Methods in molecular biology · doi:10.1007/978-1-0716-4196-5_8
Canatalay, P. J., Namazova, M., & Atalay, V. E. (2025) · ALKÜ Fen Bilimleri Dergisi · doi:10.46740/alku.1522590
Rieth, M. D. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.12.31.630938
Unlocking the Secrets of NSP3: AlphaFold2-assisted Domain Determination in SARS-CoV-2 ProteinPreprint
Edich, M., Briggs, D. C., Gao, Y., et al. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.01.14.632924
Tientcheu, L. D., Faal, F., Top, N., et al. (2025) · Research Square · doi:10.21203/rs.3.rs-5913893/v1
“Thirdly, while tools like AlphaFold provided valuable insights, their accuracy in predicting the effect of mutations on protein stability ( ΔΔ G) remains limited 41 .”
Expression and immunological role of FUNDC2 in pan-cancer
Qiu, X., Wang, S., Li, C., & Wang, Y. (2025) · PLoS ONE · doi:10.1371/journal.pone.0319343 · Cited by 1
FluxMD: High-Throughput Protein Stress Barcoding for Dynamic Structural AnalysisPreprint
Jeong, J. M. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.06.17.659767
“High-confidence AF predictions (high pLDDT scores) reflect coordinate precision, not dynamical stability[14, 15].”
Berdugo, J., Arenas, L. M. P., Montoya-Ruíz, C., & Maya, W. D. C. (2025) · Journal of Buffalo Science · doi:10.6000/1927-520x.2025.14.11
Schwabe, J., Feria, J. M., Glatter, T., Blumenkamp, P., Rupp, O., Goesmann, A., Valvano, M. A., & Søgaard‐Andersen, L. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.09.10.675344
Design of TCR-mimicking binders for pHLA with high potencyPreprint
Xie, T., Fu, Y., Gong, S.-M., Burrato, D., & Zhou, R. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.10.03.680404
A deep learning approach for rational affinity maturation of anti-VEGF nanobodiesPreprint
Verdon, G., David, L., de Brevern, A. G., & Behbahani, Y. M. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.10.20.683442
“Studies suggest AlphaFold2 [12] is not reliable for predicting mutation-induced changes in protein stability or function, as it can model folded structures for mutations that experimentally yield unfolded proteins [28, 20].”
Sarmiento, M. A. Á., & Verdesoto, E. S. B. (2025) · IEEE Ecuador Technical Chapters Meeting (ETCM) · doi:10.1109/etcm67548.2025.11304495
Unlocking the secrets of SARS-CoV-2 nsp3 by combining experiments with AlphaFold2 domain prediction
Edich, M., Briggs, D. C., Gao, Y., & Thorn, A. (2026) · Life Science Alliance · doi:10.26508/lsa.202503247 · Cited by 2
Artificial Intelligence in Human Genetics
Brandes, N. (2026) · Annual Review of Genomics and Human Genetics · doi:10.1146/annurev-genom-111124-032550 · Cited by 1
Zargari, G., Irani, M. A., Assadizadeh, M., Khoniki, S., & Eslami, M. J. (2026) · Results in Chemistry · doi:10.1016/j.rechem.2026.103190
Mancino, G., Schulz, T., Scrima, A., Müller, M. P., Bauer, S., & Rauh, D. (2026) · Journal of Medicinal Chemistry · doi:10.1021/acs.jmedchem.5c03702 · Cited by 1
Applications and insights of NMR in enzyme engineering and biocatalysis
Pérez, Y., Moreno, C. J., & Clapés, P. (2026) · Biocatalysis and Biotransformation · doi:10.1080/10242422.2026.2639009
Analysis of Protein Dynamics and Flexibility with AlphaFold
Asadi, F. (2025) · International Conference on Orange Technology (ICOT) · doi:10.1109/icot68409.2025.11425302
“A study found that AlphaFold's pLDDT confidence scores show only a weak correlation with ΔΔG values, indicating that it is ineffective for predicting the impact of mutations on protein stability [14].”
Influential
Metadata
- Authors
- Pak, M. A., Markhieva, K. A., Novikova, M. S., Petrov, D. S., Vorobyev, I. S., Maksimova, E. S., Kondrashov, F. A., & Ivankov, D. N.
- Affiliations
- Skolkovo Institute of Science and Technology (RU)
- Peoples' Friendship University of Russia (RU)
- University of New Mexico (US)
- Ural Federal University (RU)
- Institute of Science and Technology Austria (AT)
- Okinawa Institute of Science and Technology Graduate University (JP)
- Corresponding author
- Ivankov, Dmitry N.
- Venue
- PLOS One
- Publisher
- Public Library of Science (PLoS)
- Published
- 2023-03-16
- Volume
- 18
- Issue
- 3
- Pages
- e0282689
- ISSN
- 1932-6203 (ISSN-L)
- PMC
- PMC10019719
- OpenAlex
- W4327566741
- Cited by
- 322
- References
- 32
- Citations by year
- 2026: 51 · 2025: 90 · 2024: 77 · 2023: 66
- FWCI
- 25.19
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- PLoS ONE (publishedVersion, cc-by)
- bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by-nc-nd)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- RePEc: Research Papers in Economics (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Electronic Archive of Ural Federal University (ELAR UrFU) (submittedVersion, cc-by)
- Topics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Machine Learning in Bioinformatics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Structural Biology › Enzyme Structure and Function
- Keywords
- AlphaFold, single mutation, protein stability, protein folding, pLDDT, ΔΔG, protein function, green fluorescent protein, fluorescence, protein structure prediction, protein folding problem
- MeSH terms
- Amino Acid Sequence, Mutation, Proteins, Protein Folding, Protein Stability
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (qwen3.8-omni-flash).
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
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.