Investigating the volume and diversity of data needed for generalizable antibody–antigen ΔΔG prediction
Open accessThe DOI matches its Crossref record; 14 of the 14 references it cites were found in a registry or an index.
Abstract
Antibody–antigen binding affinity lies at the heart of therapeutic antibody development: efficacy is guided by specific binding and control of affinity. Here we present Graphinity, an equivariant graph neural network architecture built directly from antibody–antigen structures that achieves test Pearson’s correlations of up to 0.87 on experimental change in binding affinity (ΔΔ G ) prediction. However, our model, like previous methods, appears to be overtraining on the few hundred experimental data points available and performance is not robust to train–test cut-offs. To investigate the amount and type of data required to generalizably predict ΔΔ G , we built synthetic datasets of nearly 1 million FoldX-generated and >20,000 Rosetta Flex ddG-generated ΔΔ G values. Our results indicate that there are currently insufficient experimental data to accurately and robustly predict ΔΔ G , with orders of magnitude more likely needed. Dataset size is not the only consideration; diversity is also an important factor for model predictiveness. These findings provide a lower bound on data requirements to inform future method development and data collection efforts.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
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×
“Graphinity was not predictive on this dataset (Pearson’s correlation of 0.02), consistent with the remaining challenges in modeling antibody–antigen complexes 33 .”
InfluentialEvolutionary-scale prediction of atomic level protein structure with a language model
Lin, Z., Akin, H., Rao, R., Hie, B., Zhu, Z., Lu, W., Smetanin, N., Verkuil, R., Kabeli, O., Shmueli, Y., et al. (2022) · doi:10.1101/2022.07.20.500902 · Cited by 5,990
The FoldX web server: an online force field
Schymkowitz, J., Borg, J., Stricher, F., Nys, R., Rousseau, F., & Serrano, L. (2005) · Nucleic Acids Research · doi:10.1093/nar/gki387 · Cited by 2,759
HH-suite3 for fast remote homology detection and deep protein annotation
Steinegger, M., Meier, M., Mirdita, M., Vöhringer, H., Haunsberger, S. J., & Söding, J. (2019) · BMC Bioinformatics · doi:10.1186/s12859-019-3019-7 · Cited by 1,145
SAbDab: the structural antibody database
Dunbar, J., Krawczyk, K., Leem, J., Baker, T., Fuchs, A., Georges, G., Shi, J., & Deane, C. M. (2014) · Nucleic Acids Research · doi:10.1093/nar/gkt1043 · Cited by 674
Self-citationJankauskaitė, J., Jiménez-García, B., Dapkūnas, J., Fernández-Recio, J., & Moal, I. H. (2019) · Bioinformatics · doi:10.1093/bioinformatics/bty635 · Cited by 389 · cited 6×
“We benchmarked the performance of Flex ddG under these conditions on the Experimental_ΔΔ G _608 dataset, derived from the SKEMPI 2.0 database 31 .”
InfluentialOlsen, T. H., Boyles, F., & Deane, C. M. (2022) · Protein Science · doi:10.1002/pro.4205 · Cited by 319
Self-citationMason, D. M., Friedensohn, S., Weber, C. R., Jordi, C., Wagner, B., Meng, S. M., Ehling, R. A., Bonati, L., Dahinden, J., Gainza, P., et al. (2021) · Nature Biomedical Engineering · doi:10.1038/s41551-021-00699-9 · Cited by 308
Kovaltsuk, A., Leem, J., Kelm, S., Snowden, J., Deane, C. M., & Krawczyk, K. (2018) · The Journal of Immunology · doi:10.4049/jimmunol.1800708 · Cited by 247
Self-citationWang, M., Cang, Z., & Wei, G.-W. (2020) · Nature Machine Intelligence · doi:10.1038/s42256-020-0149-6 · Cited by 202 · cited 8×
“This value (~0.64) was lower than the Pearson’s methods are trained on features derived from antibody–antigen complex structures, such as binding surface area, interatomic interactions and energy-based terms 14,17,18 .”
Influential
Cited by
An integrated technology for quantitative wide mutational scanning of human antibody Fab libraries
Petersen, B. M., Kirby, M. B., Chrispens, K. M., Irvin, O. M., Strawn, I. K., Haas, C. M., Walker, A. M., Baumer, Z. T., Ulmer, S. A., Ayala, E., et al. (2024) · Nature Communications · doi:10.1038/s41467-024-48072-z · Cited by 24
Quast, N. P., Abanades, B., Guloglu, B., Karuppiah, V., Harper, S., Raybould, M. I. J., & Deane, C. M. (2025) · Communications Biology · doi:10.1038/s42003-025-07708-6 · Cited by 22
Self-citationHutchinson, M., Ruffolo, J. A., Haskins, N., Iannotti, M., Vozza, G., Pham, T., Mehzabeen, N., Shandilya, H., Rickert, K., Croasdale-Wood, R., et al. (2024) · mAbs · doi:10.1080/19420862.2024.2362775 · Cited by 22
Ramon, A., Ni, M., Predeina, O., Gaffey, R., Kunz, P., Onuoha, S., & Sormanni, P. (2025) · mAbs · doi:10.1080/19420862.2024.2442750 · Cited by 13
How well do contextual protein encodings learn structure, function, and evolutionary context?
(2025) · doi:10.1016/j.cels.2025.101201 · Cited by 3
Metadata
- Authors
- Hummer, A. M., Schneider, C., Chinery, L., & Deane, C. M.
- Venue
- Nature Computational Science
- Published
- 2025-07-08
- Volume
- 5
- Issue
- 8
- Pages
- 635-647
- ISSN
- 2662-8457 (ISSN-L)
- OpenAlex
- W4412102685
- Cited by
- 52
- References
- 14
- FWCI
- 11.50
- Type
- journal article
- Language
- en
- Open access
- hybrid
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Investigating the volume and diversity of data needed for generalizable antibody–antigen ΔΔG prediction. https://app.cochat.ai/papers/10.1038/s43588-025-00823-8
- The paper
- Hummer, A. M., Schneider, C., Chinery, L., & Deane, C. M. (2025). Investigating the volume and diversity of data needed for generalizable antibody–antigen ΔΔG prediction. Nature Computational Science. https://doi.org/10.1038/s43588-025-00823-8
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.