Related Experiment Video
Updated: May 26, 2026
A Simple Flow Cytometry Based Assay to Determine In Vitro Antibody Dependent Enhancement of Dengue Virus Using Zika Virus Convalescent Serum
Published on: April 10, 2018
Predicting the infecting dengue serotype from antibody titre data using machine learning.
Bethan Cracknell Daniels1, Darunee Buddhari2, Taweewun Hunsawong2
1MRC Centre for Global Infectious Disease Analysis and the Abdul Latif Jameel Institute for Disease and Emergency Analytics, School of Public Health, Imperial College London, London, United Kingdom.
Machine learning models can predict dengue virus serotypes from antibody titres, aiding vaccine development. These models help understand dengue infection history and identify protection correlates, improving clinical trial evaluations.
Area of Science:
- Virology
- Immunology
- Data Science
Background:
- Developing a universal dengue vaccine is challenging due to four serotypes and difficulty measuring specific immune responses.
- The plaque reduction neutralization test (PRNT) is standard but struggles to differentiate antibody types and infection history.
Purpose of the Study:
- To develop and validate machine learning models for predicting dengue virus serotype infection using PRNT antibody titre data.
- To assess the efficacy of machine learning in characterizing dengue infection history and identifying serotype-specific correlates of protection.
Main Methods:
- Applied four machine learning classifiers and multinomial logistic regression to PRNT antibody titres from Thai children.
- Validated models against infecting serotype identified by RT-PCR, using bootstrap sampling for performance calculation.
- Incorporated spatiotemporal data to assess its impact on predictive accuracy.
Main Results:
- Machine learning models predicted infecting serotype with up to 76.3% accuracy using PRNT data alone.
- Highest antibody titre changes did not always match the infecting serotype, especially in individuals with prior infections or Japanese encephalitis virus exposure.
- Incorporating spatiotemporal data improved prediction accuracy to 80.6%.
Conclusions:
- Machine learning classifiers effectively interpret PRNT titres, overcoming limitations in characterizing dengue infection history.
- These models are valuable tools for studying dengue immune dynamics and identifying correlates of protection.
- The findings support improved evaluation of clinical trial endpoints and dengue vaccine development.
