Deciphering a TB-related DNA methylation biomarker and constructing a TB diagnostic classifier

Area of Science:

• Epigenetics
• Molecular Diagnostics
• Infectious Diseases

Background:

• Tuberculosis (TB) diagnosis relies on methods with limitations.
• DNA methylation, an epigenetic modification, is increasingly explored as a biomarker.
• Identifying reliable TB-specific methylation patterns is crucial for diagnostic advancement.

Purpose of the Study:

• To systematically identify tuberculosis (TB)-related DNA methylation biomarkers.
• To construct and evaluate machine learning-based classifiers for TB diagnosis using these biomarkers.
• To assess the clinical utility of a novel 3-differentially methylated region (3-DMR) logistic regression classifier.

Main Methods:

• Systematic search of TB-related DNA methylation datasets.
• Identification of differentially methylated probes (DMPs) and regions (DMRs) using Limma and DMRcate.
• Construction of diagnostic classifiers via machine learning (elastic net, logistic regression).
• Performance evaluation in discovery and independent prospective cohorts.

Main Results:

• Identified 89 DMPs and 24 DMRs across four datasets.
• Selected nine and three DMRs using elastic net and logistic regression, respectively.
• Two DMRs (chr3 and chr6) were significantly differentially methylated in an independent cohort.
• The 3-DMR logistic regression classifier achieved high sensitivity (79.1%) and specificity (84.4%) in discovery sets, and 64.5% sensitivity and 90.3% specificity in the independent cohort.

Conclusions:

• DNA methylation biomarkers hold significant potential for TB diagnosis.
• The developed 3-DMR logistic regression classifier demonstrates promising diagnostic performance.
• Further validation is warranted to establish this classifier as a clinical tool for TB detection.