Skip to content
DraconiticResearch
Search Sources API
203,234 studies31 sourcesUpdated 51 min ago
Peer-reviewedCC BYBioinformatics (Oxford, England)

ProDTI: prior-guided positive-unlabeled learning for robust drug-target interaction prediction.

  • Yu Wang1
  • Haoxiang Zhang2
  • Ziye Zhu2
  • Qiang Wei3,4
  • Yun Li2
4 affiliations
  1. 1School of Artificial Intelligence, China Pharmaceutical University, Nanjing, 211198, China.
  2. 2School of Computer Science, Nanjing University of Posts and Telecommunications, Nanjing, 210023, China.
  3. 3Institute of Innovative Drug, China Pharmaceutical University, Nanjing, 211198, China.
  4. 4School of Pharmacy, China Pharmaceutical University, Nanjing, 211198, China.
Read full textPDFDOICite

Abstract

Motivation

Accurate Drug-Target Interaction (DTI) prediction is critical for drug discovery. While deep learning advances this field, the fundamental lack of verified negative interactions inevitably causes biased risk estimation. Although formulating DTI prediction as a Positive-Unlabeled (PU) learning problem is theoretically sound, directly applying PU learning to deep models suffers from early-stage optimization instability and decision boundary ambiguity.Results

We propose ProDTI, a prior-guided PU learning framework integrating domain knowledge into the PU objective. We construct a hybrid prior combining explicit biochemical similarities with implicit semantic representations from large-scale pre-trained models. This prior drives two synergistic mechanisms: soft-label distillation to stabilize the optimization trajectory, and hard-anchor regularization to explicitly penalize reliable negatives and sharpen decision boundaries. Evaluated on five benchmarks, ProDTI consistently outperforms state-of-the-art baselines, exhibiting superior robustness in challenging cold-start scenarios and providing a reliable tool for computational drug discovery.Availability and implementation

Source code and data are available at https://github.com/WangYu-AI4SCI/ProDTI.

Related

  1. Pep-PU-GAN: Positive-Unlabeled Adversarial Learning for Peptide Function Prediction
    F. Midjani, S. Hashemi, F. Z. Keshtkar +3
    bioRxivPreprintCC BY-NC-NDFull textBioinformatics & computational biologybioRxiv
  2. Comprehensive interaction profiling and machine learning prediction of bacteriophage infectivity across clinically diverse Pseudomonas aeruginosa
    Denish Piya, Avery J. C. Noonan, Hemaa Selvakumar +17
    bioRxiv (Cold Spring Harbor Laboratory)PreprintCC BY-NCFull text5 citedMicrobiologyOpenAlex
  3. An extensible deep learning framework for multi-type chromatin interaction prediction through hierarchical adaptation strategies
    Shuaibin Wang, Tong Chen, Weixu Liu +4
    PLOS Computational BiologyPeer-reviewedCC BY 4.0Full textMolecular biology & gene regulationCell biologyPLOSPMC
  4. Contribution of the Late Positive Potential to Emotional Memory: A Joint Neuro-Cognitive Approach to Memory Prediction
    R. Hellerstedt, J. Gunn, D. Talmi
    bioRxivPreprintAll rights reservedFull textNeurosciencebioRxiv
  5. ReTIF: Granularity-Aware Multitask Interaction Routing for RNA-Compound Interaction Prediction and Binding-Site Localization
    A. Chang, H. Zhu, H. Chen +6
    bioRxivPreprintCC BY-NCFull textBioinformatics & computational biologybioRxiv

Cite

BibTeXRIS
BibTeX
@article{wang2026prodti,
  title = {{ProDTI: prior-guided positive-unlabeled learning for robust drug-target interaction prediction.}},
  author = {Yu Wang and Haoxiang Zhang and Ziye Zhu and Qiang Wei and Yun Li},
  journal = {Bioinformatics (Oxford, England)},
  year = {2026},
  month = oct,
  doi = {10.1093/bioinformatics/btag712},
  url = {https://europepmc.org/article/MED/42842427},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - ProDTI: prior-guided positive-unlabeled learning for robust drug-target interaction prediction.
AU  - Yu Wang
AU  - Haoxiang Zhang
AU  - Ziye Zhu
AU  - Qiang Wei
AU  - Yun Li
PY  - 2026
DA  - 2026/10/01
JO  - Bioinformatics (Oxford, England)
DO  - 10.1093/bioinformatics/btag712
UR  - https://europepmc.org/article/MED/42842427
AB  - Motivation  Accurate Drug-Target Interaction (DTI) prediction is critical for drug discovery. While deep learning advances this field, the fundamental lack of verified negative interactions inevitably causes biased risk estimation. Although formulating DTI prediction as a Positive-Unlabeled (PU) learning problem is theoretically sound, directly applying PU learning to deep models suffers from early-stage optimization instability and decision boundary ambiguity.Results  We propose ProDTI, a prior-guided PU learning framework integrating domain knowledge into the PU objective. We construct a hybrid prior combining explicit biochemical similarities with implicit semantic representations from large-scale pre-trained models. This prior drives two synergistic mechanisms: soft-label distillation to stabilize the optimization trajectory, and hard-anchor regularization to explicitly penalize reliable negatives and sharpen decision boundaries. Evaluated on five benchmarks, ProDTI consistently outperforms state-of-the-art baselines, exhibiting superior robustness in challenging cold-start scenarios and providing a reliable tool for computational drug discovery.Availability and implementation  Source code and data are available at https://github.com/WangYu-AI4SCI/ProDTI.
SN  - 1367-4803
SN  - 1367-4811
ER  -