Related Experiment Video
Updated: May 9, 2025
Detection of Neutralization-sensitive Epitopes in Antigens Displayed on Virus-Like Particle VLP-Based Vaccines Using a Capture Assay
Published on: February 10, 2022
VirusImmu: a novel ensemble machine learning approach for viral immunogenicity prediction
Jing Li1, Zhongpeng Zhao2, ChengZheng Tai1
1Key Laboratory for Biomechanics and Mechanobiology of Ministry of Education, Beijing Advanced Innovation Centre for Biomedical Engineering, School of Engineering Medicine, School of Biological Science and Medical Engineering, Beihang University, 37 Xueyuan Road, Haidian Distirct, Beijing 100083, P. R. China.
Developing effective vaccines against viral threats is crucial. Machine learning models, particularly the novel VirusImmu approach, show strong potential for predicting viral immunogens, accelerating vaccine development.
Area of Science:
- Virology
- Immunology
- Bioinformatics
- Computational Biology
Background:
- Viral epidemics necessitate rapid vaccine development.
- Identifying protective immunogens is a critical, costly early step in vaccine design.
- Machine learning (ML) offers efficient analysis of large biological datasets like proteomes.
Purpose of the Study:
- To evaluate the immunogenicity prediction capabilities of various ML methods for viral B cell epitopes.
- To develop and validate a robust ML-based tool for predicting viral immunogens.
Main Methods:
- Curated a large dataset of known viral immunogens and non-immunogens.
- Performed cross-validation analysis on eight common ML methods, including Extreme Gradient Boosting, K Nearest Neighbours, and Random Forest.
- Developed a novel soft-voting ensemble approach named VirusImmu.
Main Results:
- Extreme Gradient Boosting, K Nearest Neighbours, and Random Forest demonstrated high predictive power.
- VirusImmu achieved powerful and stable viral immunogenicity prediction on test and external datasets, independent of protein length.
- VirusImmu successfully identified linear B cell epitopes for African Swine Fever Virus, confirmed via in vitro ELISA.
Conclusions:
- VirusImmu is a potent and stable tool for predicting viral protein immunogenicity.
- The developed ML approach significantly aids in identifying potential vaccine candidates, reducing experimental costs.
- VirusImmu holds substantial promise for accelerating vaccine development against emerging viral threats.
