Machine learning-enhanced immunopeptidomics applied to T-cell epitope discovery for COVID-19 vaccines

Area of Science:

• Immunology
• Computational Biology
• Vaccinology

Background:

• Developing T-cell-directed vaccines against SARS-CoV-2 requires identifying conserved epitopes for broad immunity.
• Emerging variants pose challenges to vaccine efficacy, necessitating continuous monitoring of viral antigen evolution.

Purpose of the Study:

• To introduce MHCvalidator, a machine learning framework, to enhance immunopeptidomics for T-cell epitope identification.
• To analyze SARS-CoV-2 proteomes for novel frameshifted antigens and track epitope conservation in vaccine candidates.

Main Methods:

• Developed MHCvalidator, a machine learning algorithm, to improve mass spectrometry-based immunopeptidomics.
• Analyzed 100,512 COVID-19 patient proteomes and utilized ribosome profiling.
• Implemented the EpiTrack pipeline to monitor global mutations of identified T-cell epitopes.

Main Results:

• MHCvalidator identified unique T-cell epitopes, including a frameshift epitope in truncated Spike antigen.
• Spike antigen truncation was observed in 0.85% of analyzed proteomes, revealing population-level frameshifted antigens.
• Most vaccine T-cell epitopes remained conserved, but one immunodominant epitope mutated in Delta and Omicron variants.

Conclusions:

• The study highlights critical SARS-CoV-2 antigenic features for T-cell vaccine design.
• Continuous adaptation and monitoring of T-cell epitopes are essential for developing effective and durable COVID-19 vaccines.