T-cell receptor specificity landscape revealed through de novo peptide design (2503.00648v1)

Abstract: T-cells play a key role in adaptive immunity by mounting specific responses against diverse pathogens. An effective binding between T-cell receptors (TCRs) and pathogen-derived peptides presented on Major Histocompatibility Complexes (MHCs) mediate an immune response. However, predicting these interactions remains challenging due to limited functional data on T-cell reactivities. Here, we introduce a computational approach to predict TCR interactions with peptides presented on MHC class I alleles, and to design novel immunogenic peptides for specified TCR-MHC complexes. Our method leverages HERMES, a structure-based, physics-guided machine learning model trained on the protein universe to predict amino acid preferences based on local structural environments. Despite no direct training on TCR-pMHC data, the implicit physical reasoning in HERMES enables us to make accurate predictions of both TCR-pMHC binding affinities and T-cell activities across diverse viral epitopes and cancer neoantigens, achieving up to 72% correlation with experimental data. Leveraging our TCR recognition model, we develop a computational protocol for de novo design of immunogenic peptides. Through experimental validation in three TCR-MHC systems targeting viral and cancer peptides, we demonstrate that our designs--with up to five substitutions from the native sequence--activate T-cells at success rates of up to 50%. Lastly, we use our generative framework to quantify the diversity of the peptide recognition landscape for various TCR-MHC complexes, offering key insights into T-cell specificity in both humans and mice. Our approach provides a platform for immunogenic peptide and neoantigen design, opening new computational paths for T-cell vaccine development against viruses and cancer.

Insights from De Novo Peptide Design for T-cell Receptor Specificity

The paper, "T-cell receptor specificity landscape revealed through de novo peptide design," details a computational framework employed to predict interactions between T-cell receptors (TCRs) and peptides presented on major histocompatibility complexes (MHCs), with an emphasis on generating novel peptides that can stimulate immune responses. The cornerstone of this method is the HERMES model, a 3D rotationally equivariant, structure-based neural network, which is guided by physics principles and trained extensively on protein data. HERMES, without direct training on TCR-pMHC complexes, successfully predicts the binding affinities and T-cell activities for diverse peptides, demonstrating a notable correlation of up to 72% with experimental data.

Leveraging this model, the researchers developed a dual protocol for de novo peptide design. The more basic, HERMES-fixed approach assumes that amino acid substitutions do not significantly alter the complex's conformation. Meanwhile, the more advanced, HERMES-relaxed method accounts for potential structural changes due to these substitutions. In experimental validations using three TCR-MHC systems, designed peptides with up to five amino acid differences from native sequences exhibited substantial T-cell activation, achieving a success rate of 50%.

The paper also explores TCR specificity by quantifying the diversity of peptides recognized by various TCR-MHC complexes. The findings indicate that a typical pair can recognize approximately 1,000 peptides, providing insights into the T-cell recognition landscape in humans and mice. This model does not only predict TCR-pMHC interactions but also suggests potential for designing neoantigen libraries for personalized cancer vaccines.

Among the remarkable outcomes are the model's predictions of T-cell activity across multiple systems, including cancer-testis antigens from NY-ESO and variants of the viral CMV antigen. The structure-based predictions achieved correlations surpassing those from sequence-based models like TULIP, especially under conditions where comprehensive structural data were available.

This paper highlights several implications:

1. Structural Impact: The reliance on structural data to guide machine learning predictions indicates a shift towards more physically-informed modeling approaches in immunology, which could address the limitations of sequence-only data.
2. Neoantigen Library Design: The capability to design peptides with improved T-cell reactivity and specificity could revolutionize personalized vaccine development, particularly for cancer, by improving the selection process for neoantigen targets.
3. Therapeutic Applications: The findings inform therapeutic strategies involving TCR-engineered cell therapies, offering a foundation to refine TCRs for enhanced specificity and reduced off-target effects.
4. Combinatorial Exploration: While current approaches focus on substituting single amino acids, future research could expand these methods to explore multi-site mutations, potentially uncovering more potent immunogenic sequences in a vast sequence space.

The paper substantiates the potential of structure-based design in immuno-engineering applications and proposes a robust computational path for creating novel peptides. Future work could improve the integration of structural and sequence-based methodologies to enhance the generalizability and applicability of such predictions across diverse antigen landscapes and immune responses.