Blood transcriptional signatures of asymptomatic FDG-PET/CT-defined Subclinical TB lesions and their prediction of future TB progression

Diagnostic tests for tuberculosis (TB) infection poorly predict future incident disease risk. We investigated a cohort of asymptomatic HIV-uninfected household contacts of TB in South Africa in which baseline lung lesions indicative of Subclinical TB identified by [18F]-fluoro-2-deoxy-D-glucose (FDG)-positron emission/computed tomography (PET/CT) had high predictive value for future TB over 5 years. RNA-sequencing analysis of whole blood samples at baseline PET/CT, 5-15m PET/CT2, and the TB diagnosis of incident TB revealed persistent enrichment of neutrophil and monocyte transcriptional modules in those with Subclinical TB and worsening and discrete lung abnormalities, while expression of a core set of type I IFN genes related to FDG-avid lymph nodes. Leveraging multiple machine learning algorithms, we derived gene signatures of various PET/CT lesion Subclinical phenotypesTB, future lesion worsening and TB diagnosis. prevalent and incident TB. Validating signatures in two independent cohorts against TB progression within 6,12,18 and fup to 24-29-months, confirmed gave AUC of 0.75-0.95 and 0.77-0.87 with performance above optimal sensitivity and wminimum specificity for the WHO target product profile for TB progression.