SETE: Sequence-based Ensemble learning approach for TCR Epitope binding prediction.

Area of Science:

• Immunology
• Bioinformatics
• Computational Biology

Background:

• T cell receptor (TCR) and epitope binding prediction is vital for immunotherapy, vaccine development, and cancer treatment.
• Current methods often rely on time-consuming feature extraction (e.g., V(D)J gene locus, amino acid biophysics) and have limited predictive performance.
• The role of k-mers (short amino acid chains) in TCR sequence-based epitope recognition remains underexplored, with unclear binding mechanisms.

Purpose of the Study:

• To develop a novel, accurate model for predicting T cell receptor (TCR) and epitope binding.
• To investigate the utility of sequence-derived k-mer features for TCR-epitope interaction prediction.
• To improve upon existing methods for predicting TCR epitope specificity.

Main Methods:

• Developed SETE (Sequence-based Ensemble learning approach for TCR Epitope binding prediction), a novel prediction model.
• Deconstructed TCR CDR3β sequences into k-mer features representing adjacent amino acid chains.
• Employed a gradient boosting decision tree algorithm to learn patterns in k-mer features for predicting TCR-epitope binding.

Main Results:

• SETE accurately predicts TCR epitope binding.
• The model effectively utilizes k-mer features derived from TCR sequences.
• SETE demonstrates superior performance compared to state-of-the-art methods on the VDJdb dataset for predicting TCR epitope specificity.

Conclusions:

• SETE offers a powerful and accurate approach for predicting TCR-epitope binding.
• Sequence-based k-mer analysis is a viable strategy for understanding TCR epitope recognition.
• The developed model can aid in the design of immunotherapies and cancer treatments by predicting TCR specificity.