Related Experiment Video
Updated: May 11, 2026
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
Predicting Viral Infection From High-Dimensional Biomarker Trajectories
Minhua Chen1, Aimee Zaas, Christopher Woods
1Minhua Chen is Ph.D. Student, Electrical and Computer Engineering Department, Aimee Zaas is Associate Professor, Christopher Woods is Associate Professor, Geoffrey S. Ginsburg is Professor and Director of Genomic Medicine, and Joseph Lucas is Assistant Research Professor, Institute for Genome Sciences and Policy & Department of Medicine, David Dunson is Professor, Department of Statistical Science, and Lawrence Carin is Professor and Department Chair ( lcarin@ee.duke.edu ), Electrical and Computer Engineering Department, Duke University, Durham, NC 27708-0291.
This study introduces a new Bayesian dynamic factor analysis method to predict infection status from gene expression data. It accurately identifies infected individuals early, even with complex, high-dimensional biomarker data.
Area of Science:
- Biostatistics
- Computational Biology
- Infectious Disease Modeling
Background:
- Predicting latent health status from high-dimensional, time-varying biomarkers is challenging.
- Time-course gene expression data from influenza challenge studies in human volunteers provides a relevant dataset.
- Existing methods may struggle with the complexity and scale of such biological data.
Purpose of the Study:
- To develop a novel time-aligned Bayesian dynamic factor analysis methodology.
- To accurately predict infection status using dynamic biomarker trajectories.
- To handle high-dimensional data where biomarkers exceed the number of individuals.
Main Methods:
- Developed a time-aligned Bayesian dynamic factor analysis framework.
- Utilized a nonparametric cure rate model for latent infection initiation times.
- Modeled gene expression trajectories as functions of low-dimensional latent factors.
Main Results:
- Accurate prediction of infected individuals before clinical symptoms manifest.
- Successful application to time-course gene expression data from influenza studies.
- Demonstrated effectiveness even when the number of biomarkers is very large compared to the number of subjects.
Conclusions:
- The proposed statistical framework enables early and accurate prediction of infection.
- The method is robust to high-dimensional data and individual variability in infection timing.
- Provides biological insights into viral response pathways.
