AFsample: improving multimer prediction with AlphaFold using massive sampling
Open accessAbstract
SUMMARY: The AlphaFold2 neural network model has revolutionized structural biology with unprecedented performance. We demonstrate that by stochastically perturbing the neural network by enabling dropout at inference combined with massive sampling, it is possible to improve the quality of the generated models. We generated ∼6000 models per target compared with 25 default for AlphaFold-Multimer, with v1 and v2 multimer network models, with and without templates, and increased the number of recycles within the network. The method was benchmarked in CASP15, and compared with AlphaFold-Multimer v2 it improved the average DockQ from 0.41 to 0.55 using identical input and was ranked at the very top in the protein assembly category when compared with all other groups participating in CASP15. The simplicity of the method should facilitate the adaptation by the field, and the method should be useful for anyone interested in modeling multimeric structures, alternate conformations, or flexible structures. AVAILABILITY AND IMPLEMENTATION: AFsample is available online at http://wallnerlab.org/AFsample.
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?
For antibody-antigen complexes, AF3 top-ranked predictions keep improving with more model seeds, even at as many as 1,000 (Wilcoxon signed-rank test between 5 and 1,000 seeds, P = 2.0 × 10−5 for percentage correct and P = 0.009 for percentage very high accuracy; ranking by protein-protein interface ipTM).
“A class of targets in which we observe this effect strongly is antibody–antigen complexes, similar to other recent work46.”
From Accurate structure prediction of biomolecular interactions with AlphaFold 3 (2024) · cites this paper as [46]
Split into 2 atoms — 0 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 1 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measuredThe ternary complex prediction is consistent with isolation of a Pol δ - Pol α -RFC complex from mammalian cells.
“it is consistent with isolation of a Pol δ - Pol α -RFC complex from mammalian cells71.”
From Predictomes, a classifier-curated database of AlphaFold-modeled protein-protein interactions (2025) · cites this paper as [71]
The Results section discusses AFsample predictions for CASP15 targets, including a nanobody interaction and a dimer, but does not address Pol δ, Pol α, RFC, or isolation of a complex from mammalian cells. Searches for the named complex and for a ternary polymerase–RFC prediction found no relevant passage in our copy. Page numbers were not available. (LLM, conf=0.77)
Experimental — accuracy not yet measuredEnabling dropout layers in the neural network, normally used only during training, is another approach to increase conformational sampling.
“Another approach is to enable dropout layers in the neural network, which are usually commonly used only during neural network training8,106,107.”
From The power and pitfalls of AlphaFold2 for structure prediction beyond rigid globular proteins (2024) · cites this paper as [106]
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
ColabFold - Making protein folding accessible to all
Mirdita, M., Schütze, K., Moriwaki, Y., Heo, L., Ovchinnikov, S., & Steinegger, M. (2021) · doi:10.1101/2021.08.15.456425 · Cited by 9,391 · see citing works
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
Scoring function for automated assessment of protein structure template quality
Zhang, Y., & Skolnick, J. (2004) · Proteins: Structure, Function, and Bioinformatics · doi:10.1002/prot.20264 · Cited by 2,392 · see citing works
UniRef clusters: a comprehensive and scalable alternative for improving sequence similarity searches
Suzek, B. E., Wang, Y., Huang, H., McGarvey, P. B., Wu, C. H., & the UniProt Consortium (2015) · Bioinformatics · doi:10.1093/bioinformatics/btu739 · Cited by 1,928 · see citing works
Mariani, V., Biasini, M., Barbato, A., & Schwede, T. (2013) · Bioinformatics · doi:10.1093/bioinformatics/btt473 · Cited by 1,138 · see citing works
Improved prediction of protein-protein interactions using AlphaFold2
Bryant, P., Pozzati, G., & Elofsson, A. (2021) · doi:10.1101/2021.09.15.460468 · Cited by 1,025 · see citing works
Clustering huge protein sequence sets in linear time
Steinegger, M., & Söding, J. (2018) · Nature Communications · doi:10.1038/s41467-018-04964-5 · Cited by 875 · see citing works
Uniclust databases of clustered and deeply annotated protein sequences and alignments
Mirdita, M., von den Driesch, L., Galiez, C., Martin, M. J., Söding, J., & Steinegger, M. (2016) · Nucleic Acids Research · doi:10.1093/nar/gkw1081 · Cited by 710 · see citing works
Antibody Structure and Function: The Basis for Engineering Therapeutics
Chiu, M. L., Goulet, D. R., Teplyakov, A., & Gilliland, G. L. (2019) · Antibodies · doi:10.3390/antib8040055 · Cited by 646 · see citing works
DockQ: A Quality Measure for Protein-Protein Docking Models
Basu, S., & Wallner, B. (2016) · PLOS ONE · doi:10.1371/journal.pone.0161879 · Cited by 460 · see citing works
Self-citationAlphaFold2 and the future of structural biology
Cramer, P. (2021) · Nature Structural & Molecular Biology · doi:10.1038/s41594-021-00650-1 · Cited by 325 · see citing works
State-of-the-Art Estimation of Protein Model Accuracy Using AlphaFold
Roney, J. P., & Ovchinnikov, S. (2022) · Physical Review Letters · doi:10.1103/physrevlett.129.238101 · Cited by 281 · see citing works
Improving peptide-protein docking with AlphaFold-Multimer using forced sampling
Johansson-Åkhe, I., & Wallner, B. (2022) · Frontiers in Bioinformatics · doi:10.3389/fbinf.2022.959160 · Cited by 162 · see citing works
Self-citationImproved multimer prediction using massive sampling with <scp>AlphaFold</scp> in <scp>CASP15</scp>
Wallner, B. (2023) · Proteins: Structure, Function, and Bioinformatics · doi:10.1002/prot.26562 · Cited by 53 · see citing works
Self-citationImproved Multimer Prediction using Massive Sampling with AlphaFold in CASP15
Wallner, B. (2023) · doi:10.22541/au.168170986.60158084/v1 · Cited by 53 · see citing works
Self-citation
Flagged first, then by citations
Research works printed but not linked: 2
- Gal Y, Ghahramani Z. Dropout as a Bayesian approximation: representing model uncertainty in deep learning. PMLR 2016;48: 1050– 1059.
- Wayment-Steele HK, Ovchinnikov S, Colwell L et al Prediction of multiple conformational states by combining sequence clustering with AlphaFold2. bioRxiv, https://doi.org/2022.10.17.512570.
Cited by
Sorted by citations
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 · see citing works
“A class of targets in which we observe this effect strongly is antibody–antigen complexes, similar to other recent work 46 .”
ResultHigh-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
“Another approach is to enable dropout layers in the neural network, which are usually commonly used only during neural network training 8,106,107 .”
BackgroundFrom interaction networks to interfaces, scanning intrinsically disordered regions using AlphaFold2
Bret, H., Gao, J., Zea, D. J., Andreani, J., & Guerois, R. (2024) · Nature Communications · doi:10.1038/s41467-023-44288-7 · Cited by 103 · see citing works
Predictomes, a classifier-curated database of AlphaFold-modeled protein-protein interactions
Schmid, E. W., & Walter, J. C. (2025) · Molecular Cell · doi:10.1016/j.molcel.2025.01.034 · Cited by 120 · see citing works
“Some of these can probably be eliminated by folding pairs in all 5 models, increased sampling 74 , or segmenting proteins 75,76 .”
Wu, D., Yin, R., Chen, G., Ribeiro-Filho, H. V., Cheung, M., Robbins, P. F., Mariuzza, R. A., & Pierce, B. G. (2024) · Science Advances · doi:10.1126/sciadv.adq6150 · Cited by 27 · 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
Physical-aware model accuracy estimation for protein complex using deep learning method
Wang, H., Sun, M., Xie, L., Liu, D., & Zhang, G. (2025) · Computational and Structural Biotechnology Journal · doi:10.1016/j.csbj.2025.01.017 · Cited by 5 · see citing works
“Since the structure of protein complexes is the key to understanding its function, methods such as AlphaFold-Multimer [2], DMFold-Multimer [3], AFsample [4], trRosettaX2 [5], and the recently released AlphaFold3[6] have been actively developed to predict the structure of multimers.”
Lalit, F., & Jose, A. M. (2025) · Nucleic Acids Research · doi:10.1093/nar/gkae1246 · Cited by 3 · see citing works
Liu, L., Liu, Y., Ye, X., Zhang, S., Li, Y., Zhu, K., Xue, Y., Zhou, J., Zhang, X., & Fang, X. (2026) · Nature Machine Intelligence · doi:10.1038/s42256-026-01264-2 · Cited by 1 · see citing works
- Authors
- Wallner, B.
- Affiliations
- Linköping University (SE)
- Corresponding author
- Wallner, Björn
- ORCID
- Wallner, Björn
- Venue
- Bioinformatics
- Publisher
- Oxford University Press (OUP)
- Published
- 2023-09-01
- Volume
- 39
- Issue
- 9
- ISSN
- 1367-4803 (ISSN-L)
- 1367-4811
- PMC
- PMC10534052
- OpenAlex
- W4386793933
- Cited by
- 164
- References
- 16
- Citations by year
- 2026: 35 · 2025: 65 · 2024: 56 · 2023: 8
- FWCI
- 14.70
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Bioinformatics (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DiVA (Linkoping University) (submittedVersion, other-oa)
- Funders
- Wallenberg AI, Autonomous System and Software Program
- Swedish e-Science Research Center
- 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
- AlphaFold-Multimer, protein structure prediction, CASP15, protein assembly, template-based modeling, alternate conformations, flexible structures
- MeSH terms
- Protein Conformation, Proteins, Neural Networks, Computer, Benchmarking
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
