Related Experiment Video
Updated: Aug 6, 2026
08:48
Generating De Novo Antigen-specific Human T Cell Receptors by Retroviral Transduction of Centric Hemichain
Published on: October 25, 2016
DynaTCR: dynamic hard-negative ensemble graph learning improves TCR-epitope binding prediction
Xiangzheng Fu1, Xinyu Zhang2, Linlin Zhuo2
1Faculty of Computer Science and Artificial Intelligence, Shenzhen University of Advanced Technology, Shenzhen 518107, China.
Bioinformatics (Oxford, England)
|July 21, 2026
Summary
DynaTCR, a novel framework, enhances T-cell receptor-epitope binding prediction using dynamic graph ensembles. It overcomes data limitations and model smoothing for improved immunotherapy development.
Area of Science:
- Immunology
- Computational Biology
- Bioinformatics
Background:
- T-cell receptors (TCRs) are crucial for adaptive immunity, recognizing peptide-MHC complexes.
- Accurate computational prediction of TCR-epitope binding (TEB) is vital for immunotherapy but faces challenges like limited data and model over-smoothing.
Purpose of the Study:
- To develop an advanced computational framework for predicting TCR-epitope binding (TEB).
- To address limitations in existing methods, including data scarcity and graph model over-smoothing.
Main Methods:
- Introduced DynaTCR, a dynamic graph ensemble learning framework.
- Utilized protein language model embeddings for TCR and epitope sequences.
- Employed a graph regularization-variance-preserving aggregation (GR-VPA) encoder and global attention layer.
- Implemented iterative hard-negative sample updates to mitigate false-negative predictions.
Main Results:
- DynaTCR demonstrated significant AUC improvements (4.0-8.2 percentage points) over existing methods on public datasets.
- Achieved superior AUPR (up to 15.8 percentage points) and an AUC of 95.1% on a curated dataset.
- Showcased high performance on an independent structure-derived test set (AUC 72.6%), confirming robustness.
Conclusions:
- DynaTCR offers a robust and effective approach for TCR-epitope binding prediction.
- The framework shows promise for accelerating immunotherapy development through accurate candidate prioritization.
