Network-guided transcriptomics identifies drug-linked host-response biomarkers in tuberculosis disease-state...

Area of Science:

• Immunology
• Genomics
• Infectious Diseases

Background:

• Distinguishing active tuberculosis (ATB) from the post-therapy state is challenging due to broad host-response signatures.
• There is a need for robust, interpretable markers to consistently track this transition across different datasets.

Purpose of the Study:

• To define a focused and biologically interpretable transcriptional signature for differentiating active tuberculosis (ATB_pre) from the post-therapy state (ATB_12m).
• To validate this signature in independent datasets and assess its utility in disease monitoring.

Main Methods:

• Integrative analysis of blood gene expression data from discovery and validation cohorts.
• Differential expression analysis, network-guided gene prioritization, and external validation.
• Machine learning models (Random Forest, Support Vector Machine) for classification performance assessment.

Main Results:

• A compact eight-gene panel, dominated by interferon-associated genes (STAT1, GBP1), was identified.
• The panel showed consistent expression changes between ATB_pre and ATB_12m across cohorts (95.9% consistent direction).
• Machine learning models achieved high accuracy (ROC-AUC 0.969) in distinguishing disease states, with STAT1 and GBP1 as key features.

Conclusions:

• The identified eight-gene panel effectively reflects transcriptional shifts between active tuberculosis and the post-therapy state.
• This signature, particularly STAT1 and GBP1, offers a promising tool for future evaluation in tuberculosis disease-state assessment and monitoring.