Abstract
<title>Abstract</title> <p> Background. Blood-based host transcriptional signatures are promising tools for tuberculosis (TB) diagnosis and triage, but individual microarray studies are limited by small sample sizes, platform differences, and cohort composition. We performed a random-effects meta-analysis of public blood transcriptomic datasets to define the reproducible core of the host response to active TB. Methods. Six Gene Expression Omnibus datasets spanning five microarray platforms and cohorts from Europe, Africa, and Asia were harmonized (363 samples after quality control: 151 active TB, 59 latent TB, 153 healthy controls). Per-dataset differential expression (active TB versus healthy) was computed with moderated t-tests, converted to Hedges' g standardized effect sizes, and pooled using DerSimonian-Laird random-effects meta-analysis, with a fixed-effect model as sensitivity analysis. Genes with Benjamini-Hochberg false discovery rate (FDR) < 0.05 and |g| ≥ 0.5 that were measured in at least four of six datasets formed the consensus signature. Enrichment analysis used Enrichr (Gene Ontology Biological Process, KEGG, Reactome). Results. Of 18,302 genes tested, 2,938 consensus differentially expressed genes were identified (1,166 up-regulated, 1,772 down-regulated). The most strongly up-regulated genes were <italic>BATF2</italic> (g = 2.67), <italic>GBP5</italic> (g = 2.50), <italic>ANKRD22</italic> (g = 2.49), <italic>FCGR1A</italic> (g = 2.46), and <italic>TNFSF10</italic> (g = 2.38); <italic>PIK3IP1</italic> was the most strongly down-regulated (g = − 1.78). Up-regulated genes converged on innate immunity: interferon signalling, neutrophil degranulation (145 genes, FDR = 6.7 × 10 − 64), NOD-like receptor signalling, phagosome, and the KEGG tuberculosis pathway itself (37 genes, FDR = 8.3 × 10 − 10). Down-regulated genes converged on cytoplasmic translation, ribosome biogenesis, and T-cell biology. Between-study heterogeneity was substantial (median I² = 54.3%), yet 86.6% of consensus genes were recovered by the fixed-effect model. No gene reproducibly distinguished latent TB from healthy controls across the three datasets containing latent-TB samples (4,300 genes tested; minimum FDR = 0.60). Conclusions. Active TB drives a highly reproducible blood transcriptional program of innate immune activation—interferon signalling, guanylate-binding proteins, Fc-γ receptors, complement, and neutrophil degranulation—coupled to suppression of lymphocyte and translational machinery. Latent TB lacks a comparably robust blood signature at the effect sizes considered here. </p>