A Two-Gene Signature for Tuberculosis Diagnosis in Persons With Advanced HIV

Abstract

Background: Transcriptomic signatures for tuberculosis (TB) have been proposed and represent a promising diagnostic tool. Data remain limited in persons with advanced HIV.Methods: We enrolled 30 patients with advanced HIV (CD4 <100 cells/mm3) in India; 16 with active TB and 14 without. Whole-blood RNA sequencing was performed; these data were merged with a publicly available dataset from Uganda (n = 33; 18 with TB and 15 without). Transcriptomic profiling and machine learning algorithms identified an optimal gene signature for TB classification. Receiver operating characteristic analysis was used to assess performance.Results: Among 565 differentially expressed genes identified for TB, 40 were shared across India and Uganda cohorts. Common upregulated pathways reflect Toll-like receptor cascades and neutrophil degranulation. The machine-learning decision-tree algorithm selected gene expression values from RAB20 and INSL3 as most informative for TB classification. The signature accurately classified TB in discovery cohorts (India AUC 0.95 and Uganda AUC 1.0; p < 0.001); accuracy was fair in external validation cohorts.Conclusions: Expression values of RAB20 and INSL3 genes in peripheral blood compose a biosignature that accurately classified TB status among patients with advanced HIV in two geographically distinct cohorts. The functional analysis suggests pathways previously reported in TB pathogenesis.

A multi-center, prospective cohort study of whole blood gene expression in the tuberculosis-diabetes interaction

“…The detailed performance of each signature in each clinical conditions is shown in the Supplementary File S2 . Moreover, linear modeling allows the comparison of performance from signatures composed by DEGs and composed by scores, as well as a fair comparison between them 46 , 47 . The outcomes were binarized to measure the sensitivity and specificity of classification, allowing us to measure each group rate and plot each signature AUC and CI value, from each country.…”

An integrative multi-omics approach to characterize interactions between tuberculosis and diabetes mellitus

“…This algorithm has been used by our group in TB and HIV coinfection studies. 25 , 26 This approach was also applied to the conceptualization of predictive models in cardiovascular diseases 27 and to predict outcomes in neurosurgery. 28 After cross-validation, MMP-28, LTE-4, 11- dTxB2, PGDM, FBXO6, SECTM1, and LINCO2009 were selected.…”

A three‐methylation‐driven gene–based deep learning model for tuberculosis diagnosis in patients with and without human immunodeficiency virus co‐infection

“…For example, Dawany et al 9 developed a 251‐gene support vector machine model that classified patients with HIV/TB co‐infection and patients with HIV monoinfection with a sensitivity of 76.2% and a specificity of 86.4%. Kulkarni et al 10 fitted a two‐gene decision tree model that distinguished HIV/TB co‐infection from HIV monoinfection among patients in the training cohorts with area under the curve (AUC) ranking between 0.95 and 1.00 and among patients in the validated cohorts with AUC ranking between 0.682 and 0.748. While the overall performance of these models is encouraging, improvements are still required before they can actually work effectively.…”