Naive and memory T cells TCR-HLA-binding prediction.

Area of Science:

• Immunology
• Bioinformatics
• Computational Biology

Background:

• T cells recognize antigens via T cell receptor (TCR) and peptide-MHC interactions.
• Peripheral T cells are expected to preferentially bind host MHC alleles.
• Predicting TCR-MHC binding is essential for understanding T cell repertoire selection.

Purpose of the Study:

• To develop a computational method for predicting TCR-MHC binding independent of peptide.
• To assess TCR-MHC binding preferences within T cell repertoires.
• To create a robust tool for analyzing TCR-HLA interactions in both naive and memory T cells.

Main Methods:

• Utilized Natural Language Processing to train a TCR-MHC binding classifier.
• Developed a two-stage prediction model, CLAIRE, accounting for multiple host HLA alleles.
• Created a CD4-CD8 T cell classifier for bulk sequencing data analysis.

Main Results:

• Initial classifier achieved high AUC (>0.90) but accuracy dropped on repertoires.
• CLAIRE demonstrated improved precision for memory T cells and cross-dataset transferability.
• The CD4-CD8 classifier achieved high AUC (0.96 and 0.90) on large datasets.

Conclusions:

• CLAIRE accurately predicts TCR-HLA binding across different T cell types and datasets.
• The tool enhances our ability to study T cell repertoire selection and function.
• CLAIRE is publicly available for research use.