Deep peptide recognition profiling decodes TCR specificity and enables disease-associated antigen discovery.

Area of Science:

• Immunology
• Computational Biology
• Structural Biology

Background:

• Predicting T cell receptor (TCR) specificity from sequence is difficult due to low correlation between sequence similarity and antigen recognition.
• Existing methods struggle to capture the nuances of TCR-peptide interactions.

Purpose of the Study:

• To develop a system for generating deep peptide recognition profiles (PRPs) for individual TCRs.
• To improve the prediction of TCR specificity and T cell activation.
• To discover novel autoantigens and understand TCR recognition mechanisms.

Main Methods:

• Integration of high-throughput yeast display with fine-tuned protein language models (pLMs).
• Generation of detailed PRPs for HLA-B*27:05-restricted TCRs from patients with ankylosing spondylitis and acute anterior uveitis.
• Training pLMs on PRPs to predict T cell activation and comparing performance against AlphaFold3 and tFold-TCR.

Main Results:

• The developed system generates comprehensive PRPs, revealing that HLA-B*27:05-restricted TCRs primarily recognize peptides via CDR3β.
• pLMs trained on PRPs significantly outperform AlphaFold3 and tFold-TCR in predicting T cell activation.
• Novel candidate autoantigens were discovered and validated.
• Model generalization was found to correlate with functional distance (PRP divergence), not sequence similarity.
• A model-intrinsic uncertainty metric was introduced for quantifying prediction confidence.

Conclusions:

• The integrated system offers a scalable method for mapping TCR recognition, significantly advancing antigen discovery.
• This approach accelerates the identification of disease-associated antigens and provides a foundation for rational TCR engineering.
• The developed PRPs and predictive models enhance our understanding of TCR-antigen interactions and T cell responses.