Computational design of class II MHC binding peptide with sequence-based evolution information

Area of Science:

• Immunology and computational biology
• Molecular modeling and drug design

Background:

• Major Histocompatibility Complex class II (MHCII)-peptide binding is essential for adaptive immunity, initiating T cell responses.
• Designing artificial MHCII-binding peptides is vital for vaccine development but challenging due to peptide flexibility.

Purpose of the Study:

• To develop a novel computational method for designing artificial MHCII-binding peptides.
• To leverage sequence-based evolutionary information and advanced structure prediction for peptide design.

Main Methods:

• Trained a Transformer neural network using sequence-based evolutionary data from native peptides.
• Incorporated amino acid frequency distributions and joint frequency distributions from multiple sequence alignments.
• Utilized an accurate sequence-based scoring function and AlphaFold3 for structure prediction.

Main Results:

• Designed artificial peptides predicted to have binding affinities comparable to native peptides.
• Achieved high structural confidence (pLDDT > 90.0) for designed peptides bound to MHCII.
• Demonstrated the potential of sequence-based methods for functional peptide design.

Conclusions:

• Established a novel paradigm for designing functional peptides using AI and sequence-based data.
• The developed method offers significant assistance for biomedical researchers in vaccine development and other fields.
• Highlights the power of integrating evolutionary information with deep learning for peptide design.