Cross-validation of existing signatures and derivation of a novel 29-gene transcriptomic signature predictive of progression to TB in a Brazilian cohort of household contacts of pulmonary TB. (January 2020)

• Record Type: Journal Article
• Authors: Leong, Samantha; Zhao, Yue; Ribeiro-Rodrigues, Rodrigo; Jones-López, Edward C.; Acuña-Villaorduña, Carlos; Rodrigues, Patricia Marques; Palaci, Moises; Alland, David; Dietze, Reynaldo; Ellner, Jerrold J.; Johnson, W. Evan; Salgame, Padmini
• Is Part Of: Tuberculosis. Volume 120(2020)
• Publication Date: 2020-01
• Subjects: Tuberculosis -- Biomarkers -- Transcriptional signatures -- TB risk signature; 616.995
• DOI: 10.1016/j.tube.2020.101898
• Languages: English
• ISSNs: 1472-9792

Abstract

Abstract: The goal of this study was to identify individuals at risk of progression and reactivation among household contacts (HHC) of pulmonary TB cases in Vitoria, Brazil. We first evaluated the predictive performance of six published signatures on the transcriptional dataset obtained from peripheral blood mononuclear cell samples from HHC that either progressed to TB disease or not (non-progressors) during a five-year follow-up. The area under the curve (AUC) values for the six signatures ranged from 0.670 to 0.461, and the PPVs did not reach the WHO published target product profiles (TPPs). We therefore used as training cohort the earliest time-point samples from the African cohort of adolescents (GSE79362) and applied an ensemble feature selection pipeline to derive a novel 29-gene signature (PREDICT29). PREDICT29 was tested on 16 progressors and 21 non-progressors. PREDICT29 performed better in segregating progressors from non-progressors in the Brazil cohort with the area under the curve (AUC) value of 0.911 and PPV of 20%. This proof of concept study demonstrates that PREDICT29 can predict risk of progression/reactivation to clinical TB disease in recently exposed individuals at least 5 years prior to disease development. Upon validation in larger and geographically diverse cohorts, PREDICT29 can be used to risk-stratify recently infected for targeted therapy.