ANTIPASTI: Interpretable prediction of antibody binding affinity exploiting normal modes and deep learning

Area of Science:

• Immunology
• Computational Biology
• Structural Biology

Background:

• High antibody binding affinity is crucial for immune responses and therapeutic applications.
• Predicting antibody-antigen interactions accurately remains a challenge.

Purpose of the Study:

• To develop a novel computational model for predicting antibody binding affinity.
• To leverage structural and energetic features for enhanced prediction accuracy.

Main Methods:

• A convolutional neural network (CNN) model named ANTIPASTI was developed.
• Input features are normal mode correlation maps from elastic network models, representing antibody-antigen structures.
• The model captures structural and energetic patterns of residue fluctuations.

Main Results:

• ANTIPASTI achieved state-of-the-art performance in predicting antibody binding affinity.
• Learned representations are interpretable, revealing binding pattern similarities and quantifying the contribution of antibody regions.
• Results highlight the significance of the antigen's influence on normal mode dynamics and cooperative effects in binding.

Conclusions:

• ANTIPASTI provides an accurate and interpretable method for predicting antibody binding affinity.
• The model's success underscores the importance of considering both structural and dynamic energetic features.
• This approach can advance antibody engineering and drug discovery efforts.