Predicting viral exposure response from modeling the changes of co-expression networks using time series gene...

Area of Science:

• Genomics
• Systems Biology
• Bioinformatics

Background:

• Understanding disease mechanisms requires linking gene expression to clinical outcomes.
• Current methods often use static gene data, missing dynamic biological processes.
• Time-series gene expression data holds potential for predicting clinical response but is challenging to analyze.

Purpose of the Study:

• To develop a statistical framework for analyzing time-series gene expression data.
• To predict clinical exposure response by capturing dynamic co-expression network changes.
• To reveal biological insights into disease development and physiological processes.

Main Methods:

• A novel statistical framework integrating co-expression network dynamics and time-series gene expression data.
• Change point detection to identify shifts in co-expression networks.
• Bayesian multiple kernel learning for exposure response prediction.
• Utilizing informative gene sets and kernel functions for subject similarity measurement.

Main Results:

• The proposed method demonstrates superior explanatory power and classification accuracy compared to existing machine learning algorithms on simulated data.
• Successfully predicted early-stage exposure response (within 24 hours) in subjects exposed to respiratory viruses.
• Identified gene sets enriched in pathways relevant to respiratory and influenza virus infections.

Conclusions:

• Dynamic changes in biological systems are linked to clinical responses.
• Analyzing inter-gene relationships within gene sets can yield novel biological insights, especially when single-gene or gene-set relationships are not significant.