pMHChat, characterizing the interactions between major histocompatibility complex class II molecules and peptides...

Area of Science:

• Immunology
• Computational Biology
• Bioinformatics

Background:

• Understanding major histocompatibility complex (MHC) class II-peptide interactions is vital for immune system research.
• Applications include neoantigen design, vaccine development, and personalized immunotherapy.

Purpose of the Study:

• To develop a predictive model for MHC class II-peptide binding reactivity, affinity, and residue contact profiling.
• To enhance the accuracy and provide detailed insights into peptide-MHC (pMHC) complex interactions.

Main Methods:

• Integration of large language models (LLMs) and deep hypergraph learning.
• A four-stage process: LLM fine-tuning, feature encoding/map fusion, task-specific prediction, and downstream analysis.
• Utilized pMHChat model with MHC pseudo-sequences and peptide sequences as input.

Main Results:

• Achieved superior performance in binding reactivity prediction (AUC: 0.8744, AUPRC: 0.8390) via five-fold cross-validation.
• Demonstrated strong binding affinity prediction (Pearson correlation: 0.7311).
• Exhibited top performance in leave-one-molecule-out and independent evaluations, with residue contact profiling capabilities.

Conclusions:

• pMHChat significantly advances predictive accuracy for MHC-peptide binding.
• The model offers valuable residue contact profiling for understanding critical binding patterns.
• pMHChat is a powerful tool for immunological research and therapeutic development.