Attention-aware differential learning for predicting peptide-MHC class I binding and T cell receptor recognition

Area of Science:

• Computational immunology
• Bioinformatics
• Machine learning in immunology

Background:

• Neoantigen identification is critical for developing effective cancer vaccines, diagnostics, and immunotherapies.
• Accurately modeling major histocompatibility complex class I (MHC-I) presentation of peptides and T cell receptor (TCR) recognition of peptide-MHC-I (pMHC-I) complexes is a significant computational challenge.
• Existing prediction methods face limitations due to complex binding motifs and sparse data in public databases.

Purpose of the Study:

• To develop advanced computational models for predicting pMHC-I binding and TCR-pMHC-I complex recognition.
• To improve the accuracy and generalization capabilities of neoantigen prediction tools.
• To identify key molecular features driving pMHC-I and TCR interactions.

Main Methods:

• Proposed an attention-aware deep learning framework with two components: TranspMHC for pMHC-I binding prediction and TransTCR for TCR-pMHC-I recognition prediction.
• TranspMHC utilizes attention mechanisms for enhanced prediction accuracy.
• TransTCR incorporates transfer learning and a differential learning strategy for improved performance and generalization.

Main Results:

• TranspMHC demonstrated superior performance over existing algorithms on independent datasets at both pan-specific and allele-specific levels.
• TransTCR exhibited enhanced generalization and superior performance on independent datasets compared to current methods.
• Identified key amino acids involved in peptide binding motifs and TCR recognition, suggesting model interpretability.

Conclusions:

• The proposed TranspMHC and TransTCR models offer significant advancements in predicting pMHC-I binding and TCR recognition.
• These models have the potential to accelerate the development of personalized cancer vaccines and immunotherapies.
• The identified key amino acids provide insights into the molecular basis of immune recognition.