OmniTCR: a foundation model unifying T cell receptor recognition prediction and conditional sequence generation
- 1West China Biomedical Big Data Center, West China Hospital, Sichuan University
Abstract
T cell receptor (TCR) recognition prediction and receptor generation are traditionally modelled separately, leaving vast TCR sequence collections disconnected from smaller TCR-peptide-MHC datasets. Here we present OmniTCR, a 113-million-parameter autoregressive foundation model pretrained on 328 million formatted human immune-sequence records. Sequence-type tokens and complementary component orders enable joint learning from individual TCR chains and partial or complete TCR-pMHC associations. On unseen epitopes, OmniTCR achieved AUPRCs of 0.7009 for peptide-TCR{beta}; recognition and 0.8235 for TCR-pMHC interaction prediction, exceeding the strongest evaluated comparators by 0.3396 and 0.3451, respectively. It distinguishes cancer from healthy repertoires across 11 independent pan-cancer cohorts (mean AUROC, 0.9436). The model achieved the highest sequence recovery on internal and external generation benchmarks. Structural modelling supported the plausibility of selected pMHC-conditioned CDR3{beta}; candidates. OmniTCR bridges heterogeneous immune sequence data, providing a foundation for computational immunology and receptor design.
Related
- Unifying Repbase and Dfam: a new open foundation for transposable element research
- Driver state detection and recognition in conditional automated driving: A review
- Probing EEG Foundation Model Attention Maps for Lightweight Interpretable Emotion Recognition
- trFlow: ultrafast all-atom protein conformational ensemble generation via conditional flow matching
- PerturbBridge: Conditional Latent Schr\"{o}dinger Bridge for Single-Cell Perturbation Prediction
BibTeX
@misc{zeng2026omnitcr,
title = {{OmniTCR: a foundation model unifying T cell receptor recognition prediction and conditional sequence generation}},
author = {F. Zeng and D. Feng and D. Song and L. Ding and Z. Tan and Q. Lei and W. Lei and A.-Y. Guo},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.10.750588},
url = {https://www.biorxiv.org/content/10.64898/2026.09.10.750588v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - OmniTCR: a foundation model unifying T cell receptor recognition prediction and conditional sequence generation
AU - F. Zeng
AU - D. Feng
AU - D. Song
AU - L. Ding
AU - Z. Tan
AU - Q. Lei
AU - W. Lei
AU - A.-Y. Guo
PY - 2026
DA - 2026/09/10
PB - bioRxiv
DO - 10.64898/2026.09.10.750588
UR - https://www.biorxiv.org/content/10.64898/2026.09.10.750588v1
AB - T cell receptor (TCR) recognition prediction and receptor generation are traditionally modelled separately, leaving vast TCR sequence collections disconnected from smaller TCR-peptide-MHC datasets. Here we present OmniTCR, a 113-million-parameter autoregressive foundation model pretrained on 328 million formatted human immune-sequence records. Sequence-type tokens and complementary component orders enable joint learning from individual TCR chains and partial or complete TCR-pMHC associations. On unseen epitopes, OmniTCR achieved AUPRCs of 0.7009 for peptide-TCR{beta}; recognition and 0.8235 for TCR-pMHC interaction prediction, exceeding the strongest evaluated comparators by 0.3396 and 0.3451, respectively. It distinguishes cancer from healthy repertoires across 11 independent pan-cancer cohorts (mean AUROC, 0.9436). The model achieved the highest sequence recovery on internal and external generation benchmarks. Structural modelling supported the plausibility of selected pMHC-conditioned CDR3{beta}; candidates. OmniTCR bridges heterogeneous immune sequence data, providing a foundation for computational immunology and receptor design.
ER -