{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,22]],"date-time":"2026-09-22T05:01:23Z","timestamp":1790053283519,"version":"4.0.1"},"reference-count":44,"publisher":"MDPI AG","issue":"18","license":[{"start":{"date-parts":[[2026,9,14]],"date-time":"2026-09-14T00:00:00Z","timestamp":1789344000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"Jilin Province Science and Technology Department","award":["YDZJ202201ZYTS400"],"award-info":[{"award-number":["YDZJ202201ZYTS400"]}],"id":[{"id":"https:\/\/ror.org\/049x38272","id-type":"ROR","asserted-by":"crossref"}]},{"award":["YDZJ202201ZYTS400"],"award-info":[{"award-number":["YDZJ202201ZYTS400"]}],"id":[{"id":"https:\/\/ror.org\/049x38272","id-type":"ROR","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Molecules"],"abstract":"<jats:p>Accurate prediction of critical temperature, pressure, and volume is essential for thermodynamic modeling and process safety design, yet remains challenging for complex molecules under data-scarce conditions. Here, we develop a multimodal GNN-BERT framework that integrates SMILES-based chemical semantics with two-dimensional topology and three-dimensional molecular geometry for critical property prediction. BERT captures molecular sequence information, while graph neural networks learn topology- and geometry-aware representations through message passing. Evaluation on 913 chemical compounds demonstrates that the proposed framework consistently outperforms conventional machine-learning models and single-modality baselines. Importantly, comparative analyses among BERT, BERT+2D-GNN, and BERT+3D-GNN reveal that incorporating three-dimensional molecular geometry provides a consistent 5\u201310% improvement across critical temperature, pressure, and volume prediction. Additional validation using random forest and support vector regression further confirms that the predictive contribution of 3D molecular information is not architecture-dependent. These results highlight three-dimensional molecular geometry as an important structural parameter for data-driven critical property prediction and provide a reliable computational strategy for thermodynamic estimation and chemical process safety applications.<\/jats:p>","DOI":"10.3390\/molecules31183251","type":"journal-article","created":{"date-parts":[[2026,9,14]],"date-time":"2026-09-14T13:05:26Z","timestamp":1789391126000},"page":"3251","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Integrating Molecular Semantics and Three-Dimensional Geometry for Critical Property Prediction: A Multimodal GNN-BERT Framework"],"prefix":"10.3390","volume":"31","author":[{"ORCID":"https:\/\/orcid.org\/0009-0001-8226-9972","authenticated-orcid":false,"given":"Beibei","family":"Wang","sequence":"first","affiliation":[{"name":"School of Emergency Management of Jilin Province, Changchun Institute of Technology, No. 3066 Tongzhi Street, Changchun 130021, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhuoyao","family":"Lv","sequence":"additional","affiliation":[{"name":"School of Emergency Management of Jilin Province, Changchun Institute of Technology, No. 3066 Tongzhi Street, Changchun 130021, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nan","family":"Ning","sequence":"additional","affiliation":[{"name":"School of Emergency Management of Jilin Province, Changchun Institute of Technology, No. 3066 Tongzhi Street, Changchun 130021, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yichen","family":"Zhang","sequence":"additional","affiliation":[{"name":"School of Emergency Management of Jilin Province, Changchun Institute of Technology, No. 3066 Tongzhi Street, Changchun 130021, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jiquan","family":"Zhang","sequence":"additional","affiliation":[{"name":"School of Environment, Northeast Normal University, Changchun 130117, China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1968","published-online":{"date-parts":[[2026,9,14]]},"reference":[{"key":"ref_1","first-page":"1","article-title":"Progress in Research on Intrinsically Safe Synthesis Technologies for Aniline and Toluene-2,5-diamine","volume":"41","author":"Ding","year":"2024","journal-title":"Chem. Ind. Eng."},{"key":"ref_2","doi-asserted-by":"crossref","first-page":"108344","DOI":"10.1016\/j.jmgm.2022.108344","article-title":"MolRoPE-BERT: An enhanced molecular representation with Rotary Position Embedding for molecular property prediction","volume":"118","author":"Liu","year":"2022","journal-title":"J. Mol. Graph. Modell."},{"key":"ref_3","unstructured":"Lee, I., and Nam, H. (2022). Infusing Linguistic Knowledge of SMILES into Chemical Language Models. arXiv."},{"key":"ref_4","doi-asserted-by":"crossref","first-page":"132","DOI":"10.1186\/s13321-025-01062-9","article-title":"AI-powered prediction of critical properties and boiling points: A hybrid ensemble learning and QSPR approach","volume":"17","author":"Bounaceur","year":"2025","journal-title":"J. Cheminf."},{"key":"ref_5","unstructured":"Yu, L., Hao, B., and Chen, X. (2005). Edge Miracles: Phase Transitions and Critical Phenomena, Science Press."},{"key":"ref_6","doi-asserted-by":"crossref","first-page":"2757","DOI":"10.1021\/ie0008068","article-title":"A Fuzzy ARTMAP-Based Quantitative Structure Property Relationship (QSPR) for Predicting Physical Properties of Organic Compounds","volume":"40","author":"Espinosa","year":"2001","journal-title":"Ind. Eng. Chem. Res."},{"key":"ref_7","first-page":"15","article-title":"QSPR Modeling for Critical Temperatures of Organic Compounds Using Hybrid Optimal Descriptors","volume":"4","author":"Bouhedjar","year":"2019","journal-title":"Int. J. Quant. Struct.-Prop. Relat."},{"key":"ref_8","doi-asserted-by":"crossref","first-page":"199","DOI":"10.1016\/j.psep.2020.10.046","article-title":"Inherently safer design protocol for process hazard analysis","volume":"149","author":"Hastie","year":"2021","journal-title":"Process Saf. Environ. Prot. Trans. Inst. Chem. Eng. Part B"},{"key":"ref_9","doi-asserted-by":"crossref","first-page":"695","DOI":"10.1063\/1.555734","article-title":"The Homogeneous Nucleation Limits of Liquids","volume":"14","author":"Avedisian","year":"1985","journal-title":"J. Phys. Chem. Ref. Data"},{"key":"ref_10","doi-asserted-by":"crossref","first-page":"584","DOI":"10.1021\/ed027p584.3","article-title":"Introduction to chemical engineering thermodynamics","volume":"27","author":"Smith","year":"1950","journal-title":"J. Chem. Educ."},{"key":"ref_11","doi-asserted-by":"crossref","first-page":"56","DOI":"10.1016\/j.jlp.2015.01.022","article-title":"Flash points measurements and prediction for binary miscible mixtures","volume":"34","author":"Pan","year":"2015","journal-title":"J. Loss Prev. Process Ind."},{"key":"ref_12","doi-asserted-by":"crossref","first-page":"254","DOI":"10.1016\/j.jlp.2015.09.017","article-title":"A mathematical method for predicting heat of reaction of organic peroxides","volume":"38","author":"Zhou","year":"2015","journal-title":"J. Loss Prev. Process Ind."},{"key":"ref_13","doi-asserted-by":"crossref","first-page":"2373","DOI":"10.1021\/ie1013663","article-title":"Quantitative Structure Property Relationship Studies for Predicting Dust Explosibility Characteristics (Kst, Pmax) of Organic Chemical Dusts","volume":"50","author":"Reyes","year":"2011","journal-title":"Ind. Eng. Chem. Res."},{"key":"ref_14","unstructured":"Miller, F.P., Vandome, A.F., and Mcbrewster, J. (2010). Molecular Descriptor, Alphascript Publishing."},{"key":"ref_15","doi-asserted-by":"crossref","first-page":"13345","DOI":"10.1007\/s00521-021-05961-4","article-title":"Prediction of chemical compounds properties using a deep learning model","volume":"33","author":"Galushka","year":"2021","journal-title":"Neural Comput. Appl."},{"key":"ref_16","doi-asserted-by":"crossref","first-page":"777","DOI":"10.1016\/j.molliq.2017.05.023","article-title":"Generalized models for predicting the critical properties of pure chemical compounds","volume":"240","author":"Varamesh","year":"2017","journal-title":"J. Mol. Liq."},{"key":"ref_17","doi-asserted-by":"crossref","first-page":"1177","DOI":"10.1021\/ci010323u","article-title":"A Fuzzy ARTMAP Based Quantitative Structure-Property Relationship (QSPR) for Predicting Aqueous Solubility of Organic Compounds","volume":"41","author":"Yaffe","year":"2001","journal-title":"J. Chem. Inf. Comput. Sci."},{"key":"ref_18","doi-asserted-by":"crossref","first-page":"112","DOI":"10.1016\/j.tca.2017.06.021","article-title":"Quantitative structure-property relationship (QSPR) study for predicting gas-liquid critical temperatures of organic compounds","volume":"655","author":"Zhou","year":"2017","journal-title":"Thermochim. Acta"},{"key":"ref_19","doi-asserted-by":"crossref","first-page":"217","DOI":"10.1016\/S0169-7439(02)00017-5","article-title":"Radial basis function neural network-based QSPR for the prediction of critical temperature","volume":"62","author":"Yao","year":"2002","journal-title":"Chemom. Intell. Lab. Syst."},{"key":"ref_20","doi-asserted-by":"crossref","first-page":"233","DOI":"10.1080\/00986448708960487","article-title":"Estimation of Pure-Component Properties from Group-Contributions","volume":"57","author":"Joback","year":"1987","journal-title":"Chem. Eng. Commun."},{"key":"ref_21","doi-asserted-by":"crossref","first-page":"104087","DOI":"10.1016\/j.jlp.2020.104087","article-title":"Prediction of superheat limit temperatures for fuel mixtures using quantitative structure-property relationship model","volume":"64","author":"Wang","year":"2020","journal-title":"J. Loss Prev. Process Ind."},{"key":"ref_22","doi-asserted-by":"crossref","first-page":"121228","DOI":"10.1016\/j.ces.2025.121228","article-title":"Using machine learning in QSPR to estimate the boiling and critical temperatures of pure organic compounds","volume":"309","author":"Beghour","year":"2025","journal-title":"Chem. Eng. Sci."},{"key":"ref_23","doi-asserted-by":"crossref","first-page":"61","DOI":"10.1109\/TNN.2008.2005605","article-title":"The Graph Neural Network Model","volume":"20","author":"Scarselli","year":"2009","journal-title":"IEEE Trans. Neural Netw."},{"key":"ref_24","doi-asserted-by":"crossref","unstructured":"Devlin, J., Chang, M.W., Lee, K., and Toutanova, K. (2019). BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. arXiv.","DOI":"10.18653\/v1\/N19-1423"},{"key":"ref_25","first-page":"97","article-title":"SMILES. 2. Algorithm for generation of unique SMILES notation","volume":"29","author":"Weininger","year":"1989","journal-title":"J. Chem. Inf. Model."},{"key":"ref_26","unstructured":"Christofidellis, D., Giannone, G., Born, J., Winther, O., Laino, T., and Manica, M. (2023, January 23\u201329). Unifying Molecular and Textual Representations via Multi-task Language Modelling. Proceedings of the 40th International Conference on Machine Learning, Honolulu, HI, USA."},{"key":"ref_27","doi-asserted-by":"crossref","first-page":"1256","DOI":"10.1038\/s42256-022-00580-7","article-title":"Large-scale chemical language representations capture molecular structure and properties","volume":"4","author":"Ross","year":"2022","journal-title":"Nat. Mach. Intell."},{"key":"ref_28","doi-asserted-by":"crossref","first-page":"3370","DOI":"10.1021\/acs.jcim.9b00237","article-title":"Analyzing Learned Molecular Representations for Property Prediction","volume":"59","author":"Yang","year":"2019","journal-title":"J. Chem. Inf. Model."},{"key":"ref_29","unstructured":"Klicpera, J., Gro, J., and G\u00fcnnemann, S. (2020, January 26\u201330). Directional Message Passing for Molecular Graphs. Proceedings of the 8th International Conference on Learning Representations (ICLR 2020), Addis Ababa, Ethiopia."},{"key":"ref_30","doi-asserted-by":"crossref","unstructured":"Zhou, G., Gao, Z., Ding, Q., Zheng, H., Xu, H., Wei, Z., Zhang, L., and Ke, G. (2023, January 1\u20135). Uni-Mol: A Universal 3D Molecular Representation Learning Framework. Proceedings of the Eleventh International Conference on Learning Representations (ICLR 2023), Kigali, Rwanda.","DOI":"10.26434\/chemrxiv-2022-jjm0j-v4"},{"key":"ref_31","doi-asserted-by":"crossref","first-page":"2108","DOI":"10.1021\/acscentsci.5c01285","article-title":"SynLlama: Generating Synthesizable Molecules and Their Analogs with Large Language Models","volume":"11","author":"Sun","year":"2025","journal-title":"ACS Cent. Sci."},{"key":"ref_32","doi-asserted-by":"crossref","first-page":"13201","DOI":"10.1002\/ange.201707394","article-title":"Discovery of a Novel Hedgehog Signaling Pathway Inhibitor by Cell-based Compound Discovery and Target Prediction","volume":"129","author":"Waldmann","year":"2017","journal-title":"Angew. Chem."},{"key":"ref_33","unstructured":"Barranco-Altirriba, M., W\u00fcrf, V., Manzini, E., Pauling, J.K., and Perera-Lluna, A. (2024). Smile-to-Bert: A BERT architecture trained for physicochemical properties prediction and SMILES embeddings generation. bioRxiv."},{"key":"ref_34","doi-asserted-by":"crossref","first-page":"17","DOI":"10.1186\/s13321-025-00959-9","article-title":"Positional embeddings and zero-shot learning using BERT for molecular-property prediction","volume":"17","author":"Mswahili","year":"2025","journal-title":"J. Cheminf."},{"key":"ref_35","doi-asserted-by":"crossref","first-page":"118","DOI":"10.1186\/s13321-025-01060-x","article-title":"Prediction model for chemical explosion consequences via multimodal feature fusion","volume":"17","author":"Wang","year":"2025","journal-title":"J. Cheminf."},{"key":"ref_36","doi-asserted-by":"crossref","first-page":"2177","DOI":"10.1109\/TCBB.2023.3237769","article-title":"BERT2OME: Prediction of 2\u2032-O-Methylation modifications from RNA sequence by transformer Architecture based on BERT","volume":"20","author":"Soylu","year":"2023","journal-title":"IEEE\/ACM Trans. Comput. Biol. Bioinf."},{"key":"ref_37","unstructured":"Maziarka, U., Danel, T., Mucha, S., Rataj, K., Tabor, J., and Jastrzbski, S. (2020). Molecule Attention Transformer. arXiv."},{"key":"ref_38","unstructured":"Luo, S., Chen, T., Xu, Y., Zheng, S., Liu, T.-Y., Wang, L., and He, D. (2022). One Transformer Can Understand Both 2D & 3D Molecular Data. arXiv."},{"key":"ref_39","unstructured":"Gokel, G.W. (2004). Dean\u2019s Handbook of Organic Chemistry, McGraw-Hill. [2nd ed.]."},{"key":"ref_40","doi-asserted-by":"crossref","first-page":"5438","DOI":"10.1021\/acs.jcim.5c00577","article-title":"Effective and Explainable Molecular Property Prediction by Chain-of-Thought Enabled Large Language Models and Multi-Modal Molecular Information Fusion","volume":"65","author":"Jin","year":"2025","journal-title":"J. Chem. Inf. Model."},{"key":"ref_41","unstructured":"(2025, August 31). Available online: https:\/\/pubchem.ncbi.nlm.nih.gov\/."},{"key":"ref_42","unstructured":"(2025, August 31). RDKit: Open-Source Cheminformatics. Available online: https:\/\/www.rdkit.org."},{"key":"ref_43","doi-asserted-by":"crossref","first-page":"263","DOI":"10.1002\/cem.1180080405","article-title":"New molecular descriptors for 2D and 3D structures","volume":"8","author":"Todeschini","year":"1994","journal-title":"Theory J. Chemom."},{"key":"ref_44","doi-asserted-by":"crossref","first-page":"100385","DOI":"10.1016\/j.egyai.2024.100385","article-title":"Descriptors-based machine-learning prediction of cetane number using quantitative structure\u2013property relationship","volume":"17","author":"Freitas","year":"2024","journal-title":"Energy AI"}],"container-title":["Molecules"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/1420-3049\/31\/18\/3251\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,22]],"date-time":"2026-09-22T04:11:08Z","timestamp":1790050268000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/1420-3049\/31\/18\/3251"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,14]]},"references-count":44,"journal-issue":{"issue":"18","published-online":{"date-parts":[[2026,9]]}},"alternative-id":["molecules31183251"],"URL":"https:\/\/doi.org\/10.3390\/molecules31183251","relation":{},"ISSN":["1420-3049"],"issn-type":[{"value":"1420-3049","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,14]]}}}