{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"bioRxiv"}],"indexed":{"date-parts":[[2026,7,30]],"date-time":"2026-07-30T15:56:44Z","timestamp":1785427004432,"version":"3.56.0"},"posted":{"date-parts":[[2025,11,10]]},"group-title":"Biochemistry","reference-count":63,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2025,11,10]],"date-time":"2025-11-10T00:00:00Z","timestamp":1762732800000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by-nc\/4.0\/"}],"funder":[{"award":["5R01AI141452-05"],"award-info":[{"award-number":["5R01AI141452-05"]}],"id":[{"id":"https:\/\/ror.org\/01cwqze88","id-type":"ROR","asserted-by":"publisher"}]},{"award":["R21AI174157"],"award-info":[{"award-number":["R21AI174157"]}],"id":[{"id":"https:\/\/ror.org\/01cwqze88","id-type":"ROR","asserted-by":"publisher"}]},{"award":["T32 GM145437"],"award-info":[{"award-number":["T32 GM145437"]}],"id":[{"id":"https:\/\/ror.org\/01cwqze88","id-type":"ROR","asserted-by":"publisher"}]},{"award":["2216011"],"award-info":[{"award-number":["2216011"]}],"id":[{"id":"https:\/\/ror.org\/021nxhr62","id-type":"ROR","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2025,11,10]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                <jats:p>Accurate prediction of the set of sequences compatible with a protein-protein interface is an unsolved problem in biology. While supervised sequence-based models trained directly on experimental data can predict variant effects, they often fail to generalize to significantly diverged sequences. We hypothesized that incorporating information from deep learning models of proteins (e.g., ESM, ProteinMPNN) could enhance generalizability. To test this hypothesis, we designed and experimentally screened several deep mutational libraries of the SARS-CoV-2 Spike Receptor Binding Domain (RBD) for binding to the ACE2 receptor. Our large dataset encompasses over 43,000 sequence variants, exhibiting up to 26 substitutions away from the parental RBD sequence, thus exploring a significantly expanded sequence space compared to previous studies. Baseline supervised learning with one-hot encoded sequences achieved high accuracy within training sets but poor performance on unseen libraries. Integrating pre-trained protein model embeddings (ESM2) as a feature showed modest improvement in generalization. To further enhance predictive power, we developed a graph attention network architecture that combines representations of local residue environments using protein structure graphs with long-range inter-residue correlations captured by protein language model (PLM) embeddings (GAN-PLM). By explicitly modeling residue environments, interface geometry, and sequence dependencies, our graph attention model outperformed purely sequence-based models, achieving substantially higher balanced accuracies when predicting functional ACE2-binding variants across the diverse sequence space spanned by our independent libraries. This demonstrates the potential of structure- and sequence-based features into deep learning frameworks to achieve accurate and generalizable predictions of protein interface function, with broad implications for understanding and engineering protein interactions relevant to emerging infectious diseases and therapeutic protein design.<\/jats:p>","DOI":"10.1101\/2025.11.04.686550","type":"posted-content","created":{"date-parts":[[2025,11,11]],"date-time":"2025-11-11T03:40:14Z","timestamp":1762832414000},"source":"Crossref","is-referenced-by-count":0,"title":["Graph attention with structural features improves the generalizability of identifying functional sequences at a protein interface"],"prefix":"10.64898","author":[{"given":"J.","family":"Ash","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"I. M.","family":"Francino-Urdaniz","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"S. P.","family":"Kells","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"C. N.","family":"Davis","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"T. A.","family":"Whitehead","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2255-0543","authenticated-orcid":false,"given":"S. D.","family":"Khare","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2025111205350643000_2025.11.04.686550v1.1","doi-asserted-by":"publisher","DOI":"10.1021\/bi00489a001"},{"key":"2025111205350643000_2025.11.04.686550v1.2","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-024-45621-4"},{"key":"2025111205350643000_2025.11.04.686550v1.3","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.2412787122"},{"key":"2025111205350643000_2025.11.04.686550v1.4","doi-asserted-by":"publisher","DOI":"10.1038\/S41586-024-07487-W"},{"key":"2025111205350643000_2025.11.04.686550v1.5","doi-asserted-by":"publisher","DOI":"10.1126\/science.ade2574"},{"key":"2025111205350643000_2025.11.04.686550v1.6","doi-asserted-by":"publisher","DOI":"10.1126\/science.add2187"},{"key":"2025111205350643000_2025.11.04.686550v1.7","doi-asserted-by":"publisher","DOI":"10.1016\/J.CELL.2020.08.012\/ATTACHMENT\/FE8D0017-8DE9-419E-A27E-E3764D796022\/MMC3.ZIP"},{"key":"2025111205350643000_2025.11.04.686550v1.8","first-page":"521","article-title":"Imprinted SARS-CoV-2 humoral immunity induces convergent Omicron RBD evolution","volume":"614","year":"2023","journal-title":"Nature"},{"key":"2025111205350643000_2025.11.04.686550v1.9","doi-asserted-by":"crossref","first-page":"1188","DOI":"10.1038\/s41564-021-00954-4","article-title":"SARS-CoV-2 variant prediction and antiviral drug design are enabled by RBD in vitro evolution","volume":"6","year":"2021","journal-title":"Nat. Microbiol"},{"key":"2025111205350643000_2025.11.04.686550v1.10","doi-asserted-by":"publisher","DOI":"10.1016\/j.celrep.2021.109627"},{"key":"2025111205350643000_2025.11.04.686550v1.11","doi-asserted-by":"publisher","DOI":"10.1016\/j.cell.2022.08.024"},{"key":"2025111205350643000_2025.11.04.686550v1.12","doi-asserted-by":"publisher","DOI":"10.1128\/jvi.00090-22"},{"key":"2025111205350643000_2025.11.04.686550v1.13","doi-asserted-by":"crossref","unstructured":"Shrestha, L.B. , Foster, C. , Rawlinson, W. , Tedla, N. , and Bull, R.A . (2022). Evolution of the SARS-CoV-2 omicron variants BA.1 to BA.5: Implications for immune escape and transmission. Rev. Med. Virol. 32, e2381.","DOI":"10.1002\/rmv.2381"},{"key":"2025111205350643000_2025.11.04.686550v1.14","doi-asserted-by":"crossref","first-page":"e5109","DOI":"10.1002\/pro.5109","article-title":"Complete combinatorial mutational enumeration of a protein functional site enables sequence-landscape mapping and identifies highly-mutated variants that retain activity","volume":"33","year":"2024","journal-title":"Protein Sci"},{"key":"2025111205350643000_2025.11.04.686550v1.15","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-020-2852-1"},{"key":"2025111205350643000_2025.11.04.686550v1.16","doi-asserted-by":"crossref","first-page":"3503","DOI":"10.1016\/j.csbj.2022.07.001","article-title":"Research progress of reduced amino acid alphabets in protein analysis and prediction","volume":"20","year":"2022","journal-title":"Comput. Struct. Biotechnol. J"},{"key":"2025111205350643000_2025.11.04.686550v1.17","doi-asserted-by":"crossref","first-page":"2198","DOI":"10.1002\/prot.24936","article-title":"Amino acid alphabet reduction preserves fold information contained in contact interactions in proteins","volume":"83","year":"2015","journal-title":"Proteins"},{"key":"2025111205350643000_2025.11.04.686550v1.18","doi-asserted-by":"publisher","DOI":"10.2807\/1560-7917.ES.2017.22.13.30494"},{"key":"2025111205350643000_2025.11.04.686550v1.19","doi-asserted-by":"publisher","DOI":"10.1002\/bit.28564"},{"key":"2025111205350643000_2025.11.04.686550v1.20","first-page":"2579","article-title":"Visualizing Data using t-SNE","volume":"9","year":"2008","journal-title":"Journal of Machine Learning Research"},{"key":"2025111205350643000_2025.11.04.686550v1.21","doi-asserted-by":"publisher","DOI":"10.1101\/2023.09.02.556057"},{"key":"2025111205350643000_2025.11.04.686550v1.22","doi-asserted-by":"publisher","DOI":"10.1016\/j.sbi.2021.04.007"},{"key":"2025111205350643000_2025.11.04.686550v1.23","doi-asserted-by":"publisher","DOI":"10.1002\/pro.2876"},{"key":"2025111205350643000_2025.11.04.686550v1.24","doi-asserted-by":"publisher","DOI":"10.1016\/0005-2795(75)90109-9"},{"key":"2025111205350643000_2025.11.04.686550v1.25","doi-asserted-by":"publisher","DOI":"10.1093\/protein\/gzad015"},{"key":"2025111205350643000_2025.11.04.686550v1.26","doi-asserted-by":"publisher","DOI":"10.1101\/2022.04.10.487779"},{"key":"2025111205350643000_2025.11.04.686550v1.27","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-024-07966-0"},{"key":"2025111205350643000_2025.11.04.686550v1.28","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-019-12130-8"},{"key":"2025111205350643000_2025.11.04.686550v1.29","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pcbi.1005541"},{"key":"2025111205350643000_2025.11.04.686550v1.30","doi-asserted-by":"publisher","DOI":"10.1002\/pro.2897"},{"key":"2025111205350643000_2025.11.04.686550v1.31","doi-asserted-by":"crossref","first-page":"86","DOI":"10.1016\/j.cels.2018.12.004","article-title":"Epistasis in a fitness landscape defined by antibody-antigen binding free energy","volume":"8","year":"2019","journal-title":"Cell Syst"},{"key":"2025111205350643000_2025.11.04.686550v1.32","doi-asserted-by":"publisher","DOI":"10.1038\/s41589-019-0386-3"},{"key":"2025111205350643000_2025.11.04.686550v1.33","first-page":"10","article-title":"Graph attention networks","volume":"1050","year":"2017","journal-title":"stat"},{"key":"2025111205350643000_2025.11.04.686550v1.34","doi-asserted-by":"crossref","first-page":"101262","DOI":"10.1016\/j.cels.2025.101262","article-title":"Evolutionary paths that link orthogonal pairs of binding proteins","volume":"16","year":"2025","journal-title":"Cell Syst"},{"key":"2025111205350643000_2025.11.04.686550v1.35","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-018-07722-9"},{"key":"2025111205350643000_2025.11.04.686550v1.36","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btaf064"},{"key":"2025111205350643000_2025.11.04.686550v1.37","doi-asserted-by":"publisher","DOI":"10.1016\/B978-0-12-381270-4.00019-6"},{"key":"2025111205350643000_2025.11.04.686550v1.38","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-022-04586-4"},{"key":"2025111205350643000_2025.11.04.686550v1.39","doi-asserted-by":"crossref","unstructured":"Escobedo, A. , Voigt, G. , Faure, A.J. , and Lehner, B . (2025). Genetics, energetics, and allostery in proteins with randomized cores and surfaces. Science 389, eadq3948.","DOI":"10.1126\/science.adq3948"},{"key":"2025111205350643000_2025.11.04.686550v1.40","doi-asserted-by":"crossref","unstructured":"Hummer, A.M. , Schneider, C. , Chinery, L. , and Deane, C.M . (2025). Investigating the volume and diversity of data needed for generalizable antibody-antigen \u0394\u0394G prediction. Nat. Comput. Sci., 1\u201313.","DOI":"10.1038\/s43588-025-00823-8"},{"key":"2025111205350643000_2025.11.04.686550v1.41","doi-asserted-by":"publisher","DOI":"10.1093\/bib\/bbae359"},{"key":"2025111205350643000_2025.11.04.686550v1.42","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btac429"},{"key":"2025111205350643000_2025.11.04.686550v1.43","doi-asserted-by":"publisher","DOI":"10.1101\/2025.01.23.634454"},{"key":"2025111205350643000_2025.11.04.686550v1.44","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.2303590120"},{"key":"2025111205350643000_2025.11.04.686550v1.45","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-024-48072-z"},{"key":"2025111205350643000_2025.11.04.686550v1.46","doi-asserted-by":"publisher","DOI":"10.7554\/eLife.23156"},{"key":"2025111205350643000_2025.11.04.686550v1.47","doi-asserted-by":"crossref","first-page":"233","DOI":"10.1007\/978-1-4939-3569-7_14","article-title":"Generating high-accuracy peptide-binding data in high throughput with yeast surface display and SORTCERY","volume":"1414","year":"2016","journal-title":"Methods Mol. Biol"},{"key":"2025111205350643000_2025.11.04.686550v1.48","doi-asserted-by":"crossref","first-page":"2617","DOI":"10.1038\/s41467-025-57510-5","article-title":"Deep indel mutagenesis reveals the impact of amino acid insertions and deletions on protein stability and function","volume":"16","year":"2025","journal-title":"Nat. Commun"},{"key":"2025111205350643000_2025.11.04.686550v1.49","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-023-06328-6"},{"key":"2025111205350643000_2025.11.04.686550v1.50","doi-asserted-by":"crossref","unstructured":"Leonard, A.C. , Weinstein, J.J. , Steiner, P.J. , Erbse, A.H. , Fleishman, S.J. , and Whitehead, T.A . (2022). Stabilization of the SARS-CoV-2 receptor binding domain by protein core redesign and deep mutational scanning. Protein Eng. Des. Sel. 35, gzac002.","DOI":"10.1093\/protein\/gzac002"},{"key":"2025111205350643000_2025.11.04.686550v1.51","doi-asserted-by":"crossref","first-page":"e70235","DOI":"10.1002\/pro.70235","article-title":"Investigating the determinants of performance in machine learning for protein fitness prediction","volume":"34","year":"2025","journal-title":"Protein Sci"},{"key":"2025111205350643000_2025.11.04.686550v1.52","doi-asserted-by":"crossref","unstructured":"Moulana, A. , Dupic, T. , Phillips, A.M. , Chang, J. , Nieves, S. , Roffler, A.A. , Greaney, A.J. , Starr, T.N. , Bloom, J.D. , and Desai, M.M . (2022). Compensatory epistasis maintains ACE2 affinity in SARS-CoV-2 Omicron BA.1. Nat. Commun. 13, 7011.","DOI":"10.1038\/s41467-022-34506-z"},{"key":"2025111205350643000_2025.11.04.686550v1.53","doi-asserted-by":"crossref","unstructured":"Taylor, A.L. , and Starr, T.N . (2023). Deep mutational scans of XBB.1.5 and BQ.1.1 reveal ongoing epistatic drift during SARS-CoV-2 evolution. PLoS Pathog. 19, e1011901.","DOI":"10.1371\/journal.ppat.1011901"},{"key":"2025111205350643000_2025.11.04.686550v1.54","doi-asserted-by":"publisher","DOI":"10.1007\/978-1-4939-7759-8_7"},{"key":"2025111205350643000_2025.11.04.686550v1.55","doi-asserted-by":"publisher","DOI":"10.1038\/nbt0697-553"},{"key":"2025111205350643000_2025.11.04.686550v1.56","doi-asserted-by":"publisher","DOI":"10.1038\/nbt.2214"},{"key":"2025111205350643000_2025.11.04.686550v1.57","doi-asserted-by":"publisher","DOI":"10.1038\/nprot.2006.94"},{"key":"2025111205350643000_2025.11.04.686550v1.58","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pone.0118193"},{"key":"2025111205350643000_2025.11.04.686550v1.59","doi-asserted-by":"crossref","first-page":"100869","DOI":"10.1016\/j.xpro.2021.100869","article-title":"Identification of SARS-CoV-2 S RBD escape mutants using yeast screening and deep mutational scanning","volume":"2","year":"2021","journal-title":"STAR Protoc"},{"key":"2025111205350643000_2025.11.04.686550v1.60","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btr507"},{"key":"2025111205350643000_2025.11.04.686550v1.61","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btq007"},{"key":"2025111205350643000_2025.11.04.686550v1.62","doi-asserted-by":"crossref","first-page":"558247","DOI":"10.3389\/fbioe.2020.558247","article-title":"Prediction of protein mutational free energy: Benchmark and sampling improvements increase classification accuracy","volume":"8","year":"2020","journal-title":"Front. Bioeng. Biotechnol"},{"key":"2025111205350643000_2025.11.04.686550v1.63","first-page":"2825","article-title":"Scikit-learn: Machine learning in Python","volume":"12","author":"Duchesnay","year":"2011","journal-title":". the Journal of machine Learning research"}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.1101\/2025.11.04.686550","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,1,15]],"date-time":"2026-01-15T09:59:43Z","timestamp":1768471183000},"score":1,"resource":{"primary":{"URL":"http:\/\/biorxiv.org\/lookup\/doi\/10.1101\/2025.11.04.686550"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,11,10]]},"references-count":63,"URL":"https:\/\/doi.org\/10.1101\/2025.11.04.686550","relation":{},"subject":[],"published":{"date-parts":[[2025,11,10]]},"subtype":"preprint"}}