From T-cell sensitization to molecular–intelligent stratification: a roadmap for precision diagnosis of latent tuberculosis infection

One quarter of the world's population carries latent tuberculosis infection (LTBI), an invisible reservoir that must be drained to end the global tuberculosis (TB) epidemic. This review charts the evolution of LTBI diagnosis from the tuberculin skin test (TST) and interferon-gamma release assays (IGRAs) to a new molecular-intelligent paradigm. At the clinical level, we propose a CD4/age-stratified, resource-matched decision framework that delivers actionable screening and sequencing strategies for people living with HIV, children, the immunosuppressed, and pregnant women. At the biomarker level, we integrate host-derived analytes (CXCL1, CCL8, IP-10, CD38+CD27⁻ T cells, and Fc-glycosylated antibodies) with pathogen-derived antigens (dormancy survival regulator [DosR], resuscitation-promoting factor [Rpf], heparin-binding hemagglutinin [HBHA], and PE/PPE) to create a three-tier index: single-analyte triage, multi-analyte confirmation, and dynamic treatment monitoring. At the technology level, we benchmark multi-omics-plus-AI models, single-molecule Simoa arrays, and microfluidic point-of-care testing (POCT) platforms for sensitivity, accessibility, and cost. High-quality cross-population validation, standardized thresholds, and resource-tiered deployment remain the principal translational bottlenecks. We call for integrated programs that combine key-population multicenter cohorts, explainable AI, and ASSURED criteria to propel LTBI management from the T-cell-sensitization era into the molecular-intelligent age. Achieving this vision within 5 years is technically feasible and will accelerate global elimination targets by enabling precision preventive therapy at an unprecedented scale.