Related Experiment Video
Updated: May 19, 2026
09:28
Enrich and Expand Rare Antigen-specific T Cells with Magnetic Nanoparticles
Published on: November 17, 2018
AI-enabled virtual immunopeptidomics links quantitative neoantigen presentation to immunogenicity
Yuhao Tan1,2,3, Ziqi Yang2,3,4, Tong Wang2,3
1Graduate Group in Genomics and Computational Biology, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA 19104, USA.
Biorxiv : the Preprint Server for Biology
|May 18, 2026
Summary
This study introduces epiVIP, an AI tool predicting cancer neoantigen abundance from gene expression data. This method improves cancer immunotherapy by revealing a balance between neoantigen quantity and quality for effective T cell responses.
Area of Science:
- Immunology
- Computational Biology
- Genomics
Background:
- Effective anti-tumor T cell response requires both neoantigen quality (non-selfness) and quantity (abundance).
- Current neoantigen prioritization methods often neglect peptide abundance due to measurement and modeling challenges.
Purpose of the Study:
- To develop a deep learning framework, epiVIP, for predicting individual HLA-I peptide abundance using (sc)RNA-seq data.
- To investigate the relationship between neoantigen abundance and non-selfness in determining antigenicity.
- To assess the predictive value of neoantigen abundance for tumor reactivity and patient survival in cancer immunotherapy.
Main Methods:
- Developed epiVIP, a deep learning framework utilizing gene expression profiles and 1.7 million immune peptides.
- Trained and validated epiVIP on diverse clinical datasets, including 33,711 neoantigens.
- Employed T cell functional assays for experimental validation of mechanistic findings.
Main Results:
- epiVIP demonstrated strong generalizability across unseen samples.
- A compensatory relationship between neoantigen abundance and non-selfness was identified, supporting TCR avidity theory.
- Neoantigen abundance independently predicted tumor reactivity and patient survival in vaccine and immune checkpoint blockade cohorts.
- Identified PSME4's role in regulating MAGEA3 epitope presentation, validated experimentally.
Conclusions:
- AI-enabled virtual immunopeptidomics, exemplified by epiVIP, is a powerful strategy for enhancing cancer immunotherapy.
- Predicting neoantigen abundance offers crucial insights into anti-tumor immunity and patient outcomes.
- Understanding the interplay of neoantigen quantity and quality is key for optimizing immunotherapeutic strategies.
