Performance of the TB/LTBI 27 and TB/OD 44 transcript signatures and the transcript signatures of Berry et al. [25] when applied to our test cohort.

Comparison of the statistical measures of performance of disease classification using our TB/LTBI 27 and TB/OD 44 transcript signatures with the classification using the 393 (−6 transcript) and 86 (−1 transcript) transcript signatures from Berry et al. [25]. The marked improvement shown for HIV+ individuals in both TB versus LTBI and TB versus OD comparisons suggests that transcript signatures must be derived from both HIV-infected and -uninfected individuals in order to have a diagnostic value in these populations. The performance of our signatures in TB versus OD comparison highlights the need for real world “other disease” controls when deriving biomarkers from clinical cohorts.aCalculations of the differences were performed before rounding for reporting purposes on the paper.