About EpitopeGen

EpitopeGen is a breakthrough technology that maps T-cell receptor (TCR) sequences to their cognate epitope sequences using advanced machine learning.

Our model is built on a GPT-2 architecture and enhanced with semi-supervised learning to maximize training data while maintaining biological validity through distributional constraints.

The technology has been successfully applied to analyze single-cell RNA and TCR sequencing data from cancer and COVID-19 patients, identifying phenotype-associated T cells with characteristic cytotoxic markers.

Please check our preprint on BioArxiv: Repertoire-level generation of T-cell epitopes with a large-scale generative transformer

Analysis Workflow

TCR Sequences

TCR Sequences

Upload your TCR CDR3β sequences in CSV format

→
Analysis

EpitopeGen Analysis

Our transformer model predicts potential epitope sequences

→
Report

Annotation

Predicted epitopes are matched against curated databases

Try EpitopeGen

Results will be sent to this email address
[[ selectedDatabaseInfo.description ]]

The input CSV file must contain a column named 'tcr'

Service limit due to resources: 3000 TCRs

Upload

Selected file: [[ selectedFile.name ]]

Click to change file

Drop your CSV file here or click to browse

[[ statusMessage ]]

Example Results

Example EpitopeGen output visualization

After analysis, you'll receive comprehensive results that look like this:

  • tcr: Your input CDR3β sequences
  • pred_0: Top predicted epitope sequence by EpitopeGen
  • match_0: Indicator (1 = yes) if the predicted epitope was found in the selected database
  • ref_protein_0: Reference protein matching the predicted epitope
  • ref_epitope_0: Specific reference epitope from the protein that matched

When match_0 = 1 and you've selected the tumor database, this suggests your TCR may recognize tumor-associated antigens. In practice, you will get multiple predictions' results. Also, we ensemble 11 models for robustness by default.

Citation

@article{epitopegen2025,
    title={Repertoire-level generation of T-cell epitopes with a large-scale generative transformer},
    author={Minuk Ma, Wilson Tu, Carlos Vasquez-Rios, Jiarui Ding},
    journal={bioRxiv},
    year={2025},
    doi={https://doi.org/10.1101/2025.01.13.632824}
}

Contact Us

Have questions about EpitopeGen or want to collaborate? Reach out to our team.

Email minukma@cs.ubc.ca
GitHub GitHub Repository
PyPI PyPI Package