EUVIMEDEUVIMED entdecken
Crossref

CDI-DTI: A Strong Cross-Domain Interpretable Drug-Target Interaction Prediction Framework Based on Multi-Strategy Fusion

Xiangyu Li iD, Haojie Yang, Kaimiao Hu, Runzhi Wu iD, Ruibing Chen iD, Guangjian Ni, Liangliang Liu, Ran Su

Journal of Chemical Information and Modeling · 09.03.2026 · 66 (5) · 2627-2639

DOI: 10.1021/acs.jcim.5c02908 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Chemical Information and Modeling, online erschienen am 17.02.2026. Verfasst von Xiangyu Li und 7 weiteren Autor:innen. Beteiligte Einrichtung: Department of Intelligence and Computing, Tianjin University, 300072 Tianjin, China. Hauptthemen laut MEDLINE: Drug Discovery. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 53 Literaturangaben. Gefördert von: Tianjin University, Seed Foundation of Tianjin University, State Key Laboratory of Medical Proteomics.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate prediction of drug-target interaction (DTI) is pivotal for drug discovery, yet existing methods often fail to address challenges like cross-domain generalization, cold-start prediction, and interpretability. In this work, we propose CDI-DTI, a novel cross-domain interpretable framework for DTI prediction, designed to overcome these limitations. By integrating multimodal features-textual, structural, and functional-through a multistrategy fusion approach, CDI-DTI ensures robust performance across different domains and in cold-start scenarios. A multisource cross-attention mechanism is introduced to align and fuse features early, while a bidirectional cross-attention layer captures fine-grained intramodal drug-target interaction. At the late fusion stage, we incorporate Gram Loss for feature alignment and a deep orthogonal fusion module to eliminate redundancy. Experimental results on several benchmark data sets demonstrate that CDI-DTI significantly outperforms existing methods, particularly in cross-domain and cold-start tasks, while maintaining high interpretability for practical applications in drug-target interaction prediction.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Drug Discovery

Autor:innen und Institutionen

  • Department of Intelligence and Computing, Tianjin University, 300072 Tianjin, China.
  • School of Pharmaceutical Science and Technology, Faculty of Medicine, Tianjin University, 300072 Tianjin, China.
  • Academy of Medical Engineering and Translational Medicine, Tianjin University, 300072 Tianjin, China.
  • Haihe Laboratory of Brain-Computer Interaction and Human-Machine Integration, 300392 Tianjin, China.
  • State Key Laboratory of Advanced Medical Materials and Devices, 300192 Tianjin, China.
  • Tianjin Key Laboratory of Brain Science and Neuroengineering, 300072 Tianjin, China.

ORCID-Profile: Xiangyu Li, Runzhi Wu, Ruibing Chen

Förderung

  • Tianjin University
  • Seed Foundation of Tianjin University
  • State Key Laboratory of Medical Proteomics

Zitieren

APA

Li, X., Yang, H., Hu, K., Wu, R., Chen, R., Ni, G., Liu, L., & Su, R. (2026). CDI-DTI: A Strong Cross-Domain Interpretable Drug-Target Interaction Prediction Framework Based on Multi-Strategy Fusion. Journal of Chemical Information and Modeling, 66(5), 2627-2639. https://doi.org/10.1021/acs.jcim.5c02908

Vancouver

Li X, Yang H, Hu K, Wu R, Chen R, Ni G, et al. CDI-DTI: A Strong Cross-Domain Interpretable Drug-Target Interaction Prediction Framework Based on Multi-Strategy Fusion. Journal of Chemical Information and Modeling. 2026;66(5):2627-2639. doi:10.1021/acs.jcim.5c02908

BibTeX
@article{Li2026,
  author = {Li, Xiangyu and Yang, Haojie and Hu, Kaimiao and Wu, Runzhi and Chen, Ruibing and Ni, Guangjian and Liu, Liangliang and Su, Ran},
  title = {CDI-DTI: A Strong Cross-Domain Interpretable Drug-Target Interaction Prediction Framework Based on Multi-Strategy Fusion},
  journal = {Journal of Chemical Information and Modeling},
  year = {2026},
  volume = {66},
  number = {5},
  pages = {2627--2639},
  doi = {10.1021/acs.jcim.5c02908}
}
Kostenlos starten →