Tuberculosis remains a major health threat, infecting nearly a third of the world’s population. Of those infected, 5-10% progress from latent infection to active tuberculosis (TB) disease and biomarkers to identify which individuals will progress are needed to allow targeted prophylactic treatment. Several risk biomarkers have been developed to predict progression but have not been tested head-to-head on the same platform. Here, we used the NanoString platform and compared the performance of 15 published gene signatures in predicting progression at baseline in a household contact cohort. Expression of gene signatures was profiled in RNA extracted from whole blood and scored using GSVA and PLAGE. We found that specificity is enhanced by combining signatures and report that the performance of a combined signature that includes a newly derived parsimonious signature through machine learning and a published signature met WHO TPP levels for a triage test. The combined signature had a 90.9% sensitivity and 88% specificity with a PPV of 0.24 and NPV of 1. This combined signature has potential clinical utility in identifying high-risk individuals for targeted prophylaxis to prevent TB morbidity and mortality.