Detection for New Biomarkers of Tuberculosis Infection Activity Using Machine Learning Methods.

Area of Science:

• Biotechnology
• Bioinformatics
• Computational Biology

Background:

• Latent tuberculosis infection (LTBI) is a major reservoir for active tuberculosis (ATB), posing diagnostic challenges.
• Traditional immunological assays like interferon-gamma release assays (IGRAs) struggle to reliably distinguish LTBI from ATB.
• High-throughput omics technologies and machine learning (ML) offer novel approaches for precise, biomarker-based differential diagnostics.

Purpose of the Study:

• To review transcriptomic and proteomic biomarker research for discriminating LTBI from ATB.
• To emphasize machine learning-based analytical frameworks in tuberculosis diagnostics.
• To systematically compare data platforms, modeling strategies, and identify translational barriers.

Main Methods:

• Transcriptomic and proteomic profiling of host immune responses to identify LTBI and ATB signatures.
• Integration of ML techniques (feature selection, dimensionality reduction, multimodal learning, explainable AI) for robust diagnostic models.
• Utilizing single-modality (RNA-seq, microarrays, proteomics) and multimodal approaches, including deep learning frameworks.

Main Results:

• ML-driven analyses of omics data demonstrate superior sensitivity, specificity, and clinical applicability compared to conventional tests.
• Multimodal integration further enhances diagnostic accuracy and robustness.
• Advances support the development of quantitative reverse transcription PCR (qRT-PCR)-based biomarker panels for clinical application.

Conclusions:

• Machine learning and omics integration provide a promising pathway for enhancing global tuberculosis diagnostics.
• Key translational barriers include cohort homogeneity, platform dependency, and limited external validation.
• Future research should focus on improving clinical applicability through rigorous validation and addressing identified barriers.