A Highly Effective System for Predicting MHC-II Epitopes With Immunogenicity.

Area of Science:

• Immunology
• Computational Biology
• Bioinformatics

Background:

• Therapeutic cancer vaccines have shown promise in immunotherapy, but identifying effective neoantigens, particularly Major Histocompatibility Complex (MHC)-II epitopes, remains a challenge.
• Previous research focused on epitope binding or presentation, neglecting their immunogenicity, which is crucial for potent vaccine design.

Purpose of the Study:

• To develop a novel computational tool, FIONA (Flexible Immunogenicity Optimization Neural-network Architecture), for predicting both MHC-II presented epitopes and their immunogenicity.
• To improve the design of therapeutic cancer vaccines by accurately identifying CD4+ T-cell epitopes.

Main Methods:

• A convolutional neural network model (FIONA) was developed and trained on data from the IEDB (Immune Epitope Database).
• FIONA utilizes a human leukocyte antigen (HLA) allele hierarchical encoding model and peptide dense embedding fusion encoding.
• Performance was evaluated against existing tools in predicting MHC-II presented epitopes.

Main Results:

• FIONA demonstrated high accuracy in predicting MHC-II presented epitopes, achieving an Area Under the Curve (AUC) of 0.94.
• FIONA outperformed several other tools in head-to-head comparisons for MHC-II epitope prediction.
• The model uniquely incorporates the ability to predict the immunogenicity of epitopes specific to MHC-II subtypes.

Conclusions:

• FIONA provides a reliable pipeline for predicting CD4+ T-cell immune responses against cancer and infectious diseases.
• The tool facilitates the design of more effective therapeutic cancer vaccines by accurately identifying immunogenic epitopes.
• This advancement has significant implications for both cancer immunotherapy and the development of vaccines for infectious diseases.