{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,24]],"date-time":"2026-09-24T14:00:05Z","timestamp":1790258405392,"version":"4.1.0"},"reference-count":48,"publisher":"Wiley","issue":"10","license":[{"start":{"date-parts":[[2026,9,17]],"date-time":"2026-09-17T00:00:00Z","timestamp":1789603200000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2026,9,17]],"date-time":"2026-09-17T00:00:00Z","timestamp":1789603200000},"content-version":"tdm","delay-in-days":0,"URL":"http:\/\/doi.wiley.com\/10.1002\/tdm_license_1.1"}],"funder":[{"DOI":"10.13039\/100009619","name":"Japan Agency for Medical Research and Development","doi-asserted-by":"publisher","award":["JP26ama121028"],"award-info":[{"award-number":["JP26ama121028"]}],"id":[{"id":"10.13039\/100009619","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["onlinelibrary.wiley.com"],"crossmark-restriction":true},"short-container-title":["Protein Science"],"published-print":{"date-parts":[[2026,10]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>\n                    Distinguishing biological protein\u2013protein interfaces from non\u2010biological contacts remains an important task in structural biology, particularly as protein complex structures continue to accumulate. Recent advances in protein language models (pLMs) have expanded their use across diverse protein prediction tasks, including sequence\u2010based protein\u2013protein interaction prediction. Such approaches often pool representations over entire sequences, whereas structure\u2010based methods commonly rely on more complex graph\u2010 or geometry\u2010based integration. We therefore asked how much interface\u2010relevant information could be extracted by simply pooling pLM embeddings over structurally defined local regions. Here, we revisited biological\u2010versus\u2010crystal interface classification as a structurally well\u2010defined testbed for this question. We developed a framework that compares full\u2010sequence pooling, interface\u2010localized pooling of pLM embeddings, and multimodal integration of sequence\u2010derived embeddings with point\u2010cloud representations of protein surfaces. On two benchmark datasets, interface\u2010localized pooling achieved stronger performance than full\u2010sequence or non\u2010interacting surface pooling across three pLM backbones. Despite its simplicity, the resulting representation performed within the range of established methods that rely on explicit evolutionary analysis or geometric modeling. Incorporating explicit geometric surface information changed performance slightly, without reaching statistical significance. Because the model is linear, the interface\u2010level score decomposes exactly into per\u2010residue contributions, which varied within amino\u2010acid types. Together, our results indicate that the principal gain arises from localizing pLM embeddings to physically interacting residues, enabling a lightweight and accessible implementation for biological\u2010versus\u2010crystal interface classification. Code and scripts are available on GitHub:\n                    <jats:ext-link xmlns:xlink=\"http:\/\/www.w3.org\/1999\/xlink\" xlink:href=\"https:\/\/github.com\/fukasawa-group\/espma\">https:\/\/github.com\/fukasawa-group\/espma<\/jats:ext-link>\n                    .\n                  <\/jats:p>","DOI":"10.1002\/pro.70790","type":"journal-article","created":{"date-parts":[[2026,9,17]],"date-time":"2026-09-17T13:13:31Z","timestamp":1789650811000},"update-policy":"https:\/\/doi.org\/10.1002\/crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["<scp>ESpma<\/scp>\n                    : A method for assessing biological\/non\u2010biological interfaces using point\u2010clouds\u2010based structural features and protein language models"],"prefix":"10.1002","volume":"35","author":[{"given":"Sarah","family":"Nozawa","sequence":"first","affiliation":[{"name":"Center for Bioscience Research and Education Utsunomiya University  Utsunomiya Japan"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4567-4768","authenticated-orcid":false,"given":"Kentaro","family":"Tomii","sequence":"additional","affiliation":[{"name":"Artificial Intelligence Research Center National Institute of Advanced Industrial Science and Technology (AIST)  Tokyo Japan"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0882-4392","authenticated-orcid":false,"given":"Yoshinori","family":"Fukasawa","sequence":"additional","affiliation":[{"name":"Center for Bioscience Research and Education Utsunomiya University  Utsunomiya Japan"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"311","published-online":{"date-parts":[[2026,9,17]]},"reference":[{"key":"e_1_2_9_2_1","unstructured":"AkibaT SanoS YanaseT OhtaT KoyamaM.Optuna: a next\u2010generation hyperparameter optimization framework. arXiv [cs.LG] [Internet].2019Available from:http:\/\/arxiv.org\/abs\/1907.10902"},{"key":"e_1_2_9_3_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.csbj.2025.12.008"},{"key":"e_1_2_9_4_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.jmb.2003.12.073"},{"key":"e_1_2_9_5_1","doi-asserted-by":"publisher","DOI":"10.1186\/s12900-014-0022-0"},{"key":"e_1_2_9_6_1","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btn022"},{"key":"e_1_2_9_7_1","doi-asserted-by":"publisher","DOI":"10.1093\/bib\/bbae076"},{"key":"e_1_2_9_8_1","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btac020"},{"key":"e_1_2_9_9_1","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btv622"},{"key":"e_1_2_9_10_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.jmb.2024.168540"},{"key":"e_1_2_9_11_1","doi-asserted-by":"publisher","DOI":"10.1021\/acs.jcim.5b00190"},{"key":"e_1_2_9_12_1","doi-asserted-by":"publisher","DOI":"10.2307\/2531595"},{"key":"e_1_2_9_13_1","doi-asserted-by":"publisher","DOI":"10.1186\/1471-2105-13-334"},{"key":"e_1_2_9_14_1","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.061411798"},{"key":"e_1_2_9_15_1","doi-asserted-by":"crossref","unstructured":"ElnaggarA EssamH Salah\u2010EldinW MoustafaW ElkerdawyM RochereauC et al.Ankh: optimized protein language model unlocks general\u2010purpose modelling. arXiv [csLG] [Internet].2023Available from:http:\/\/arxiv.org\/abs\/2301.06568","DOI":"10.1101\/2023.01.16.524265"},{"key":"e_1_2_9_16_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-024-71783-8"},{"key":"e_1_2_9_17_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-019-48913-8"},{"key":"e_1_2_9_18_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41592-026-03011-2"},{"key":"e_1_2_9_19_1","doi-asserted-by":"publisher","DOI":"10.1074\/jbc.M309272200"},{"key":"e_1_2_9_20_1","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btu857"},{"key":"e_1_2_9_21_1","doi-asserted-by":"publisher","DOI":"10.1093\/nargab\/lqae150"},{"key":"e_1_2_9_22_1","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btaf178"},{"key":"e_1_2_9_23_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.csbj.2025.01.022"},{"key":"e_1_2_9_24_1","doi-asserted-by":"publisher","DOI":"10.1038\/nsb1297-973"},{"key":"e_1_2_9_25_1","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btz437"},{"key":"e_1_2_9_26_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-021-03819-2"},{"key":"e_1_2_9_27_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.jmb.2007.05.022"},{"key":"e_1_2_9_28_1","doi-asserted-by":"publisher","DOI":"10.1126\/science.ade2574"},{"key":"e_1_2_9_29_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-025-64512-w"},{"key":"e_1_2_9_30_1","doi-asserted-by":"publisher","DOI":"10.1186\/1471-2105-15-S16-S3"},{"key":"e_1_2_9_31_1","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btn497"},{"key":"e_1_2_9_32_1","doi-asserted-by":"publisher","DOI":"10.1038\/srep14214"},{"key":"e_1_2_9_33_1","doi-asserted-by":"publisher","DOI":"10.12688\/f1000research.7931.1"},{"key":"e_1_2_9_34_1","unstructured":"PedregosaF VaroquauxG GramfortA MichelV ThirionB GriselO et al.Scikit\u2010learn: machine learning in python. arXiv [cs.LG] [Internet]: 2825\u20132830.2012Available from:http:\/\/arxiv.org\/abs\/1201.0490"},{"key":"e_1_2_9_35_1","doi-asserted-by":"publisher","DOI":"10.1107\/S0021889803012421"},{"key":"e_1_2_9_36_1","unstructured":"QiCR YiL SuH GuibasLJ.PointNet++: deep hierarchical feature learning on point sets in a metric space. arXiv [csCV] [Internet].2017Available from:http:\/\/arxiv.org\/abs\/1706.02413"},{"key":"e_1_2_9_37_1","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btac759"},{"key":"e_1_2_9_38_1","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btaf192"},{"key":"e_1_2_9_39_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-021-27396-0"},{"key":"e_1_2_9_40_1","doi-asserted-by":"publisher","DOI":"10.1186\/1471-2105-12-77"},{"key":"e_1_2_9_41_1","doi-asserted-by":"crossref","unstructured":"SverrissonF FeydyJ CorreiaBE BronsteinMM.Fast end\u2010to\u2010end learning on protein surfaces. bioRxiv [Internet]:2020.12.28.424589.2020https:\/\/doi.org\/10.1101\/2020.12.28.424589v1.full","DOI":"10.1101\/2020.12.28.424589"},{"key":"e_1_2_9_42_1","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkae1010"},{"key":"e_1_2_9_43_1","doi-asserted-by":"publisher","DOI":"10.1002\/1097-0134(20010101)42:1<108::aid-prot110>3.0.co;2-o"},{"key":"e_1_2_9_44_1","doi-asserted-by":"publisher","DOI":"10.1038\/s42003-023-05133-1"},{"key":"e_1_2_9_45_1","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkq1059"},{"key":"e_1_2_9_46_1","doi-asserted-by":"publisher","DOI":"10.1093\/bioadv\/vbad191"},{"key":"e_1_2_9_47_1","doi-asserted-by":"publisher","DOI":"10.1007\/s10930-007-9108-x"},{"key":"e_1_2_9_48_1","doi-asserted-by":"publisher","DOI":"10.1109\/TCBB.2024.3381825"},{"key":"e_1_2_9_49_1","doi-asserted-by":"publisher","DOI":"10.1186\/1471-2105-7-27"}],"container-title":["Protein Science"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/pro.70790","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/full-xml\/10.1002\/pro.70790","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/pro.70790","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,24]],"date-time":"2026-09-24T13:08:53Z","timestamp":1790255333000},"score":1,"resource":{"primary":{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/10.1002\/pro.70790"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,17]]},"references-count":48,"journal-issue":{"issue":"10","published-print":{"date-parts":[[2026,10]]}},"alternative-id":["10.1002\/pro.70790"],"URL":"https:\/\/doi.org\/10.1002\/pro.70790","archive":["Portico"],"relation":{},"ISSN":["0961-8368","1469-896X"],"issn-type":[{"value":"0961-8368","type":"print"},{"value":"1469-896X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,17]]},"assertion":[{"value":"2026-03-26","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-08-27","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-09-17","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e70790"}}