EPIC: multi-objective guided diffusion for epitope design in TCR-pMHC complexes

Area of Science:

• Computational biology
• Immunology
• Bioinformatics

Background:

• T cell receptor (TCR) recognition of peptide-MHC (pMHC) complexes is crucial for adaptive immunity.
• Designing immunogenic epitopes is challenging due to complex binding constraints and limited data.
• Existing methods cannot simultaneously ensure antigenicity, MHC presentation, and TCR specificity.

Purpose of the Study:

• To develop a computational framework for the de novo design of epitopes.
• To address the limitations of existing epitope design methods by integrating triplet binding constraints.
• To enable the generation of epitopes with tailored immunological properties.

Main Methods:

• EPIC utilizes a multi-objective diffusion framework.
• It decomposes TCR-pMHC binding into three biologically grounded sub-tasks.
• ESM-based classifiers and a peptide diffusion generator are integrated, leveraging diverse immunological datasets.

Main Results:

• EPIC-designed epitopes show improved predicted interface energies compared to ground-truth epitopes.
• Generated epitopes demonstrate high sequence novelty (80.1%), uniqueness (100%), and diversity (64.05%).
• The framework achieves high antigenicity scores (0.4723) and comparable structural confidence.

Conclusions:

• EPIC is the first computational framework for de novo epitope design integrating TCR-pMHC triplet constraints.
• This approach facilitates the design of epitopes for personalized cancer vaccines and T cell therapies.
• EPIC represents a paradigm shift from epitope discovery to rational epitope design.