Predicting recognition between T cell receptors and epitopes with TCRGP.

Area of Science:

• Immunology
• Bioinformatics
• Computational Biology

Background:

• T cell receptors (TCRs) are crucial for adaptive immunity, recognizing pathogens and initiating immune responses.
• Analyzing TCR specificity enhances understanding of immune status in various disorders.
• Current methods for predicting TCR-epitope interactions have limitations.

Purpose of the Study:

• To develop a novel computational method, TCRGP, for predicting T cell receptor (TCR) epitope specificity.
• To leverage amino acid sequences of TCR complementarity determining regions (CDRs) for improved prediction accuracy.
• To enable advanced analysis of combined single-cell RNA and TCR sequencing data.

Main Methods:

• Developed TCRGP, a Gaussian process-based machine learning model.
• Utilized amino acid sequences from TCRα and TCRβ complementarity determining regions (CDRs).
• Evaluated TCRGP performance using epitope-specific TCR sequencing data.

Main Results:

• TCRGP demonstrated higher prediction accuracy (AUROC) compared to existing state-of-the-art methods for epitope-specificity prediction.
• The method effectively identifies important CDRs involved in epitope recognition.
• Successfully applied TCRGP to quantify epitope-specific TCRs in combined scRNA+TCRαβ sequencing data.

Conclusions:

• TCRGP offers a more accurate approach for predicting TCR-epitope interactions.
• This method advances the analysis of immune cell populations and their transcriptomic states.
• TCRGP aids in identifying disease-specific T cells, such as HBV-epitope specific T cells in hepatocellular carcinoma patients.