TCRconv: predicting recognition between T cell receptors and epitopes using contextualized motifs

Area of Science:

• Immunology
• Computational Biology
• Bioinformatics

Background:

• T cells recognize antigen fragments (epitopes) via T cell receptors (TCRs) presented by MHC molecules.
• This recognition triggers immune responses, T cell activation, and clonal expansion, forming immunological memory.
• This immunological memory has potential applications in infectious diseases, autoimmunity, and tumor immunology.

Purpose of the Study:

• To introduce TCRconv, a novel deep learning model for predicting TCR-epitope recognition.
• To leverage advanced deep learning techniques for enhanced accuracy in TCR-epitope interaction prediction.

Main Methods:

• TCRconv utilizes a deep protein language model and convolutional neural networks.
• The model extracts contextualized motifs for improved prediction accuracy.
• The model was trained and validated using TCR repertoires from COVID-19 patients.

Main Results:

• TCRconv achieves state-of-the-art accuracy in TCR-epitope prediction.
• The model provides insights into T cell dynamics and phenotypes during disease.
• Demonstrated utility in analyzing TCR repertoires from COVID-19 patients.

Conclusions:

• TCRconv is a powerful tool for predicting TCR-epitope interactions.
• The model offers valuable insights into T cell responses relevant to clinical immunology.
• TCRconv has potential applications in understanding and treating immune-related diseases.