Structure Based Prediction of Neoantigen Immunogenicity

Area of Science:

• Immunology
• Computational Biology
• Oncology

Background:

• Developing effective immunological therapies, especially personalized cancer vaccines using neoantigens, is a major goal.
• Identifying immunogenic neoantigens that trigger tumor rejection remains a significant hurdle despite advances in identification methods.

Purpose of the Study:

• To explore the potential of high-resolution structural modeling and energetic scoring for predicting neoantigen immunogenicity.
• To develop a computational method for accurately modeling peptide-MHC interactions and predicting T cell responses.

Main Methods:

• Developed a strategy for rapid and accurate modeling of nonameric peptides bound to HLA-A2.
• Trained a neural network on structural features influencing T cell receptor (TCR) and peptide binding energies.
• Evaluated the performance of the structurally-parameterized neural network against existing methods.

Main Results:

• The developed neural network accurately predicted CD8+ T cell responses for HLA-A2 presented peptides.
• The model outperformed methods lacking explicit structural or energetic information.
• The study provided insights into the structural and biophysical mechanisms of immunogenicity.

Conclusions:

• Structure-based immunogenicity prediction holds significant potential for developing personalized peptide-based cancer vaccines.
• Computational modeling and AI can enhance the identification of effective neoantigens for immunotherapy.