DeepMHCII: a novel binding core-aware deep interaction model for accurate MHC-II peptide binding affinity prediction

Area of Science:

• Immunological Bioinformatics
• Computational Biology
• Machine Learning in Immunology

Background:

• Predicting major histocompatibility complex (MHC)-peptide binding affinity is crucial in immunological bioinformatics.
• Current deep learning methods struggle with MHC class II molecules due to simplified input representations that ignore biological interaction details.

Purpose of the Study:

• To develop a deep learning model, DeepMHCII, that accurately predicts MHC class II peptide binding affinity.
• To address limitations of existing methods by integrating biological knowledge of molecular interactions.

Main Methods:

• Proposed DeepMHCII, a binding core-aware deep learning model.
• Introduced a binding interaction convolution layer to model interactions between peptide binding cores and MHC class II pseudo sequences.
• Integrated all potential binding cores within the peptide sequence.

Main Results:

• DeepMHCII significantly outperformed four state-of-the-art methods across various validation strategies.
• Demonstrated superior performance in 5-fold cross-validation, leave-one-molecule-out, and independent testing.
• Visualization confirmed the model's effectiveness and highlighted the importance of biological fact integration.

Conclusions:

• DeepMHCII achieves high predictive performance for MHC class II peptide binding affinity.
• Properly modeling biological interactions is essential for advancing deep learning in immunology.
• The model facilitates efficient knowledge discovery in the field.