TSpred: a robust prediction framework for TCR-epitope interactions using paired chain TCR sequence data.

Area of Science:

• Immunology
• Computational Biology
• Bioinformatics

Background:

• Predicting T-cell receptor (TCR)-epitope interactions is crucial for advancing cancer immunotherapy and vaccine development.
• Accurate prediction is challenging due to limited data, particularly for novel epitopes.

Purpose of the Study:

• To introduce TSpred, a deep learning model for pan-specific TCR binding specificity prediction.
• To develop a robust model capable of generalizing to unseen epitopes.

Main Methods:

• TSpred utilizes a deep learning architecture combining Convolutional Neural Networks (CNNs) and an attention mechanism.
• A reciprocal attention mechanism is employed to identify key interaction patterns between TCRs and epitopes.
• The model is trained on paired chain TCR data.

Main Results:

• TSpred achieves state-of-the-art performance in predicting TCR specificity for both known and novel epitopes.
• The model demonstrates robustness against peptide imbalance biases present in datasets.
• The reciprocal attention mechanism provides interpretability by highlighting important binding regions.

Conclusions:

• TSpred is a reliable and robust method for TCR-epitope binding prediction.
• The approach offers improved accuracy and interpretability for immunological applications.
• The developed model advances the field of predictive immunology.