T cells are a critical component of the adaptive immune system, playing a role in infectious disease, autoimmunity, and cancer. T cell function is mediated by the T cell receptor (TCR) protein, a highly diverse receptor targeting specific peptides presented by the major histocompatibility complex (pMHCs). Predicting the specificity of TCRs for their cognate pMHCs is central to understanding adaptive immunity and enabling personalized therapies. However, accurate prediction of this protein-protein interaction remains challenging due to the extreme diversity of both TCRs and pMHCs. Here, we present ImmSET (Immune Synapse Encoding Transformer), a novel sequence-based architecture designed to model interactions among sets of variable-length biological sequences. We train this model across a range of dataset sizes and compositions and study the resulting models' generalization to pMHC targets. We describe a failure mode in prior sequence-based approaches that inflates previously reported performance on this task and show that ImmSET remains robust under stricter evaluation. In systematically testing the scaling behavior of ImmSET with training data, we show that performance scales consistently with data volume across multiple data types and compares favorably with the pre-trained protein language model ESM2 fine-tuned on the same datasets. Finally, we demonstrate that ImmSET can outperform AlphaFold2 and AlphaFold3-based pipelines on TCR-pMHC specificity prediction when provided sufficient training data. This work establishes ImmSET as a scalable modeling paradigm for multi-sequence interaction problems, demonstrated in the TCR-pMHC setting but generalizable to other biological domains where high-throughput sequence-driven reasoning complements structure prediction and experimental mapping.
Figures & tables
Figure 1Figure 2Figure 3Figure 4Figure 5
HLA
n_peptides
mean
median
percentile_25
percentile_75
min
max
A*01:01
37
0.607
0.619
0.552
0.694
0.348
0.818
A*02:01
134
0.706
0.704
0.639
0.772
0.393
0.959
A*29:02
32
0.644
0.619
0.542
0.761
0.423
0.887
A*68:01
42
0.649
0.636
0.554
0.702
0.446
0.939
B*08:01
18
0.579
0.554
0.493
0.631
0.374
0.908
B*18:01
46
0.624
0.632
0.553
0.700
0.129
0.853
Table 6
Figure 7
Appendix figures & tables16 assets
Supplementary material from the paper’s appendix.
Appendix
Parameter
Value
Embedding dim.
420
Attention heads
5
Hidden dim.
420
Encoder layers
10
Dropout
0.01
CLS classifier
128
Appendix
Table 8
MLM
CSM
CLS
mean_AUROC
1
1
1
0.616
1
4
4
0.656
1
4
16
0.684
1
4
48
0.680
1
3
55
0.675
1
5
60
0.673
Appendix
Table 9
Figure 10Figure 11
Dataset
CDR 1α
CDR 2α
CDR 3α
CDR 1β
CDR 2β
CDR 3β
Train-0
6 (5–7)
7 (6–8)
13 (11–16)
5 (5–5)
6 (6–6)
14 (12–16)
Train-1
6 (5–7)
7 (6–8)
13 (11–16)
5 (5–5)
6 (6–6)
14 (12–16)
Train-2
6 (5–7)
7 (6–8)
13 (11–16)
5 (5–5)
6 (6–6)
14 (12–16)
Validation
6 (5–7)
7 (6–8)
13 (11–16)
5 (5–5)
6 (6–6)
14 (12–16)
Holdout
6 (5–7)
7 (6–8)
13 (11–16)
5 (5–5)
6 (6–6)
14 (12–16)
Appendix
Table 12
Figure 13Figure 14
Metric
Model
α (P α )
95% CI
β (T β )
95% CI
AUROC
ImmSET
0.078
(0.075, 0.082)
0.036
(0.031, 0.040)
AUROC
ESM2
0.072
(0.064, 0.079)
0.032
(0.021, 0.044)
Appendix
Table 1: We report the parameters of a best-fit power law relating AUROC to training data size and composition, quantified by P (number of peptides) and T (number of TCRs per peptide) for both ImmSET and ESM2, along with 95% confidence intervals of each scaling parameter.
Protein-protein interaction prediction uses sequence, structure, and biological networks to identify interacting proteins and their binding sites or poses.
T cell receptor (TCR)-epitope binding prediction is essential for understanding adaptive immunity and developing immunotherapies. Existing sequence- and structure-based models often generalize poorly to unseen epitopes and provide limited interpretability. Furthermore, the impact of generated structures on model learning remains unclear. We present TCR-SRIM, a structure-regularized interpretable-by-design model that combines protein language model embeddings with interpretable contact prototypes to capture residue-level TCR-epitope interactions. TCR-SRIM achieves state-of-the-art predictive performance and improved interpretation quality on the TCR-XAI benchmark. Using its inherent interpretability, we further evaluate the effect of generated structures on model learning. While structures predicted by AlphaFold3, TCRModel2, and tFold-TCR yield competitive performance, they lead to less accurate interaction patterns and reduced binding-site diversity than experimentally-resolved structures. Our results highlight limitations of current structure prediction models for TCR-epitope learning and demonstrate the value of interpretable-by-design models for studying generated biological structures.
Jiarui Li, Zixiang Yin, Yunbei Zhang +4
Department of Computer Science, Tulane University · Department of Biochemistry and Molecular Biology, Tulane University School of Medicine
We propose EMBER, a structure-based framework for TCR--pMHC binding prediction that eliminates the need for a predicted docked complex, and with it the placement error that dominates such predictions. Traditional structure-based predictors utilize the co-folded complex structure of the receptor and the antigen and interpret binding from its contact surface. Consequently, every downstream feature inherits any deviation the structure-prediction tool introduced. Methodologically, EMBER folds the TCR and pMHC independently, encodes them as SE(3)-invariant residue graphs, and compares their learned representations without constructing a shared coordinate frame. A relative pose is therefore not an input of the model, so placement error cannot propagate into the prediction. Computationally, EMBER preprocesses O(N+M) monomer folds where a complex-based predictor needs O(NM) operations. Experimentally, under epitope-disjoint evaluation with six retrained baselines, EMBER is first on every metric under all four folding tools, by 0.9 to 14.4% in micro-AUROC, 18.5 to 25.3% in micro-AUPRC and 10.6 to 24.1% in MCC. Its ranking is also up to 85% more stable across folding tools.
Accurate computational prediction of T cell receptor (TCR) antigen specificity would transform the study of T cell biology and enable scalable immune engineering, yet existing models lack sufficient sensitivity and specificity for broad applications. A major limitation is the absence of rigorously defined, unseen benchmark datasets that allow unbiased evaluation of model performance and generalizability. Here, we describe two complementary classes of datasets that meet this criterion and argue that they provide both a robust framework for model assessment and a foundation for next-generation TCR-antigen prediction algorithm development.
Yiming Liao, Yiheng Li, Ning Jiang +2
Trustworthy and Intelligent Computing Lab (TAIC), Department of Computer Science and Electrical Engineering, University of Maryland, Baltimore County, Baltimore, Maryland 21250, USA · Children’s Hospital of Philadelphia, Philadelphia, PA 19104, USA · Department of Bioengineering, University of Pennsylvania, Philadelphia, PA 19104, USA +5