{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T01:45:58Z","timestamp":1782783958579,"version":"3.54.5"},"reference-count":35,"publisher":"MDPI AG","issue":"13","license":[{"start":{"date-parts":[[2026,6,29]],"date-time":"2026-06-29T00:00:00Z","timestamp":1782691200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IJMS"],"abstract":"<jats:p>Accurate prediction of protein\u2013protein interaction (PPI) interface residues is essential for understanding molecular recognition and supporting structure-guided design. To integrate contextual sequence representations with structure-related physicochemical information, we propose a multimodal framework termed MD-Transformer. The model combines residue-level ProtBERT embeddings with physicochemical descriptors, including B-factor, solvent-accessible surface area (SASA), and hydrophobicity. A hybrid fusion module first aligns heterogeneous features, followed by Transformer encoding and cross-modal attention for multimodal integration. Using the DB5.5 benchmark, physicochemical descriptors were Z-score normalized exclusively with training-set statistics. Under the complex-level split protocol (Official A), MD-Transformer achieved an AUPRC of 0.564, outperforming the ablation model without physicochemical descriptors by 0.159 and reducing false-positive predictions on exposed non-interface residues. Under the homology-aware split protocol (Official B v1), the model maintained an AUPRC of 0.480 and an MCC of 0.242, indicating retained predictive capability under reduced sequence similarity constraints. Under the same aligned evaluation workflow, PeSTo achieved an AUPRC of 0.264. Further SASA-stratified analyses identified SASA as a major contributor to suppressing false-positive predictions across residue exposure environments, while also revealing a precision-recall trade-off in highly exposed residues. These results suggest that contextual sequence representations and residue-level physicochemical descriptors provide complementary predictive signals.<\/jats:p>","DOI":"10.3390\/ijms27135848","type":"journal-article","created":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T00:55:11Z","timestamp":1782780911000},"page":"5848","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["MD-Transformer: Multimodal Integration of ProtBERT Embeddings and Physicochemical Descriptors for Protein\u2013Protein Interface Residue Prediction"],"prefix":"10.3390","volume":"27","author":[{"given":"Jiahui","family":"Yang","sequence":"first","affiliation":[{"name":"School of Electrical and information Technology, Yunnan Minzu University, Kunming 650500, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jihua","family":"Feng","sequence":"additional","affiliation":[{"name":"School of Electrical and information Technology, Yunnan Minzu University, Kunming 650500, China"},{"name":"Yunnan Key Laboratory of Unmanned Autonomous Systems, Kunming 650500, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yuting","family":"Zhang","sequence":"additional","affiliation":[{"name":"School of Electrical and information Technology, Yunnan Minzu University, Kunming 650500, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhongxing","family":"Chen","sequence":"additional","affiliation":[{"name":"School of Electrical and information Technology, Yunnan Minzu University, Kunming 650500, China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1968","published-online":{"date-parts":[[2026,6,29]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"bbab036","DOI":"10.1093\/bib\/bbab036","article-title":"A survey on computational models for predicting protein\u2013protein interactions","volume":"22","author":"Hu","year":"2021","journal-title":"Brief. Bioinform."},{"key":"ref_2","doi-asserted-by":"crossref","first-page":"527","DOI":"10.1016\/j.tibs.2023.03.003","article-title":"Recent advances in predicting and modeling protein\u2013protein interactions","volume":"48","author":"Durham","year":"2023","journal-title":"Trends Biochem. Sci."},{"key":"ref_3","doi-asserted-by":"crossref","first-page":"1341","DOI":"10.1007\/s12551-022-01032-7","article-title":"Protein\u2013protein interaction prediction methods: From docking-based to AI-based approaches","volume":"14","author":"Tsuchiya","year":"2022","journal-title":"Biophys. Rev."},{"key":"ref_4","doi-asserted-by":"crossref","first-page":"1761","DOI":"10.1038\/s41598-020-80900-2","article-title":"Classification and prediction of protein\u2013protein interaction interface using machine learning algorithm","volume":"11","author":"Das","year":"2021","journal-title":"Sci. Rep."},{"key":"ref_5","doi-asserted-by":"crossref","first-page":"102775","DOI":"10.1016\/j.sbi.2024.102775","article-title":"Revolutionizing protein\u2013protein interaction prediction with deep learning","volume":"85","author":"Zhang","year":"2024","journal-title":"Curr. Opin. Struct. Biol."},{"key":"ref_6","doi-asserted-by":"crossref","first-page":"3386","DOI":"10.1093\/bioinformatics\/btm434","article-title":"meta-PPISP: A meta web server for protein-protein interaction site prediction","volume":"23","author":"Qin","year":"2007","journal-title":"Bioinformatics"},{"key":"ref_7","doi-asserted-by":"crossref","first-page":"e2016239118","DOI":"10.1073\/pnas.2016239118","article-title":"Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences","volume":"118","author":"Rives","year":"2021","journal-title":"Proc. Natl. Acad. Sci. USA"},{"key":"ref_8","doi-asserted-by":"crossref","first-page":"1123","DOI":"10.1126\/science.ade2574","article-title":"Evolutionary-scale prediction of atomic-level protein structure with a language model","volume":"379","author":"Lin","year":"2023","journal-title":"Science"},{"key":"ref_9","doi-asserted-by":"crossref","first-page":"7112","DOI":"10.1109\/TPAMI.2021.3095381","article-title":"ProtTrans: Toward understanding the language of life through self-supervised learning","volume":"44","author":"Elnaggar","year":"2021","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell."},{"key":"ref_10","doi-asserted-by":"crossref","first-page":"nwaf028","DOI":"10.1093\/nsr\/nwaf028","article-title":"Foundation models in bioinformatics","volume":"12","author":"Guo","year":"2025","journal-title":"Natl. Sci. Rev."},{"key":"ref_11","doi-asserted-by":"crossref","first-page":"730","DOI":"10.1038\/s41592-022-01490-7","article-title":"ScanNet: An interpretable geometric deep learning model for structure-based protein binding site prediction","volume":"19","author":"Tubiana","year":"2022","journal-title":"Nat. Methods"},{"key":"ref_12","doi-asserted-by":"crossref","first-page":"btac759","DOI":"10.1093\/bioinformatics\/btac759","article-title":"DeepRank-GNN: A graph neural network framework to learn patterns in protein\u2013protein interfaces","volume":"39","author":"Renaud","year":"2023","journal-title":"Bioinformatics"},{"key":"ref_13","doi-asserted-by":"crossref","first-page":"106762","DOI":"10.1016\/j.bpc.2022.106762","article-title":"Importance of interface and surface areas in protein-protein binding affinity prediction: A machine learning analysis based on linear regression and artificial neural network","volume":"283","author":"Yang","year":"2022","journal-title":"Biophys. Chem."},{"key":"ref_14","doi-asserted-by":"crossref","first-page":"12113","DOI":"10.1109\/TPAMI.2023.3275156","article-title":"Multimodal learning with transformers: A survey","volume":"45","author":"Xu","year":"2023","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell."},{"key":"ref_15","doi-asserted-by":"crossref","unstructured":"Tsai, Y.H.H., Bai, S., Liang, P.P., Kolter, J.Z., Morency, L.-P., and Salakhutdinov, R. (2019). Multimodal transformer for unaligned multimodal language sequences. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, Association for Computational Linguistics.","DOI":"10.18653\/v1\/P19-1656"},{"key":"ref_16","doi-asserted-by":"crossref","first-page":"558","DOI":"10.1038\/280558a0","article-title":"Temperature-dependent X-ray diffraction as a probe of protein structural dynamics","volume":"280","author":"Frauenfelder","year":"1979","journal-title":"Nature"},{"key":"ref_17","doi-asserted-by":"crossref","first-page":"1626","DOI":"10.1021\/acs.chemrev.8b00290","article-title":"Utility of B-factors in protein science: Interpreting rigidity, flexibility, and internal motion and engineering thermostability","volume":"119","author":"Sun","year":"2019","journal-title":"Chem. Rev."},{"key":"ref_18","doi-asserted-by":"crossref","first-page":"189","DOI":"10.12688\/f1000research.7931.1","article-title":"FreeSASA: An open source C library for solvent accessible surface area calculations","volume":"5","author":"Mitternacht","year":"2016","journal-title":"F1000Research"},{"key":"ref_19","doi-asserted-by":"crossref","first-page":"105","DOI":"10.1016\/0022-2836(82)90515-0","article-title":"A simple method for displaying the hydropathic character of a protein","volume":"157","author":"Kyte","year":"1982","journal-title":"J. Mol. Biol."},{"key":"ref_20","first-page":"5999","article-title":"Attention is all you need","volume":"30","author":"Vaswani","year":"2017","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"ref_21","doi-asserted-by":"crossref","first-page":"1709","DOI":"10.1109\/TCBB.2024.3410696","article-title":"CrossPredGO: A novel light-weight cross-modal multi-attention framework for protein function prediction","volume":"21","author":"Kumar","year":"2024","journal-title":"IEEE\/ACM Trans. Comput. Biol. Bioinform."},{"key":"ref_22","doi-asserted-by":"crossref","first-page":"1444","DOI":"10.1038\/s41592-024-02362-y","article-title":"Guiding questions to avoid data leakage in biological machine learning applications","volume":"21","author":"Bernett","year":"2024","journal-title":"Nat. Methods"},{"key":"ref_23","doi-asserted-by":"crossref","first-page":"3031","DOI":"10.1016\/j.jmb.2015.07.016","article-title":"Updates to the integrated protein\u2013protein interaction benchmarks: Docking benchmark version 5 and affinity benchmark version 2","volume":"427","author":"Vreven","year":"2015","journal-title":"J. Mol. Biol."},{"key":"ref_24","doi-asserted-by":"crossref","first-page":"3150","DOI":"10.1093\/bioinformatics\/bts565","article-title":"CD-HIT: Accelerated for clustering the next-generation sequencing data","volume":"28","author":"Fu","year":"2012","journal-title":"Bioinformatics"},{"key":"ref_25","doi-asserted-by":"crossref","first-page":"1026","DOI":"10.1038\/nbt.3988","article-title":"MMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets","volume":"35","author":"Steinegger","year":"2017","journal-title":"Nat. Biotechnol."},{"key":"ref_26","first-page":"82","article-title":"Pymol: An open-source molecular graphics tool","volume":"40","author":"DeLano","year":"2002","journal-title":"CCP4 Newsl. Protein Crystallogr."},{"key":"ref_27","doi-asserted-by":"crossref","first-page":"133","DOI":"10.1006\/jmbi.1997.1233","article-title":"Prediction of protein-protein interaction sites using patch analysis","volume":"272","author":"Jones","year":"1997","journal-title":"J. Mol. Biol."},{"key":"ref_28","doi-asserted-by":"crossref","first-page":"100804","DOI":"10.1016\/j.patter.2023.100804","article-title":"Leakage and the reproducibility crisis in machine-learning-based science","volume":"4","author":"Kapoor","year":"2023","journal-title":"Patterns"},{"key":"ref_29","first-page":"e70002","article-title":"Reproducibility in machine-learning-based research: Overview, barriers, and drivers","volume":"46","author":"Semmelrock","year":"2025","journal-title":"AI Mag."},{"key":"ref_30","doi-asserted-by":"crossref","first-page":"0240","DOI":"10.34133\/research.0240","article-title":"A transformer-based ensemble framework for the prediction of protein\u2013protein interaction sites","volume":"6","author":"Mou","year":"2023","journal-title":"Research"},{"key":"ref_31","unstructured":"Loshchilov, I., and Hutter, F. (2017). Decoupled weight decay regularization. arXiv."},{"key":"ref_32","doi-asserted-by":"crossref","first-page":"2175","DOI":"10.1038\/s41467-023-37701-8","article-title":"PeSTo: Parameter-free geometric deep learning for accurate prediction of protein binding interfaces","volume":"14","author":"Krapp","year":"2023","journal-title":"Nat. Commun."},{"key":"ref_33","doi-asserted-by":"crossref","first-page":"4","DOI":"10.1186\/s13040-023-00322-4","article-title":"The Matthews correlation coefficient (MCC) should replace the ROC AUC as the standard metric for assessing binary classification","volume":"16","author":"Chicco","year":"2023","journal-title":"BioData Min."},{"key":"ref_34","doi-asserted-by":"crossref","first-page":"13","DOI":"10.1186\/s13040-021-00244-z","article-title":"The Matthews correlation coefficient (MCC) is more reliable than balanced accuracy, bookmaker informedness, and markedness in two-class confusion matrix evaluation","volume":"14","author":"Chicco","year":"2021","journal-title":"BioData Min."},{"key":"ref_35","doi-asserted-by":"crossref","first-page":"44102","DOI":"10.52202\/079017-1400","article-title":"A closer look at AUROC and AUPRC under class imbalance","volume":"37","author":"McDermott","year":"2024","journal-title":"Adv. Neural Inf. Process. Syst."}],"container-title":["International Journal of Molecular Sciences"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/1422-0067\/27\/13\/5848\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T01:14:23Z","timestamp":1782782063000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/1422-0067\/27\/13\/5848"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6,29]]},"references-count":35,"journal-issue":{"issue":"13","published-online":{"date-parts":[[2026,7]]}},"alternative-id":["ijms27135848"],"URL":"https:\/\/doi.org\/10.3390\/ijms27135848","relation":{},"ISSN":["1422-0067"],"issn-type":[{"value":"1422-0067","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,6,29]]}}}