Circulating Cell-Free RNA in Blood as a Host Response Biomarker for the Detection of Tuberculosis [training_data]

Tuberculosis (TB) remains a leading cause of death from an infectious disease worldwide. This is partly due to a lack of tools to effectively screen and triage individuals with potential TB. Whole blood RNA signatures have been extensively studied as potential biomarkers for TB, but they have failed to meet the World Health Organization's (WHOs) optimal target product profiles (TPPs) for a non-sputum triage or diagnostic test. In this study, we investigated the utility of plasma cell-free RNA (cfRNA) as a host response biomarker for TB. We used RNA profiling by sequencing to analyze plasma samples from individuals with a cough lasting at least two weeks, who were seen at outpatient TB clinics in Uganda, Vietnam, and the Philippines. We split the samples into a discovery set for model training and testing, and reserved a validation set from a separate cohort to validate the model performance. We trained 15 machine learning classification models and developed a 6-gene signature that has a high performance in differentiating TB positive and negative individuals (Area Under the Curve, AUC = 0.95, 0.92, 0.95 for the test, training and validation sets respectively). This 6-gene signature exceeds the optimal WHO TPPs for a TB triage test (sensitivity: 97.1% [95% CI: 80.9-100%], specificity: 85.2% [95% CI: 72.4-100%]) and was robust to differences in geographic location, sample collection, and HIV status. Analysis of matched whole blood samples from the validation cohort highlighted the differences in origin of plasma and whole blood RNA. Overall, our results demonstrate the utility of plasma cfRNA for the detection of TB and suggest the potential for a point-of-care, gene expression-based assay to aid in early detection of TB. Overall design: In the study presented here 251 patients who presented with a cough >= 2 weeks were identified in three separate cohorts, across three different countries (Uganda, Vietnam, Philippines). Patients were identified as TB positive or negative at clinics (tb status indicated in the “tb” column of the metadata). We extracted the cell-free RNA from the plasma of patients and conducted RNA sequencing. We split these samples into discovery (cohort 1 and 2) and validation (cohort 3). And then further split the discovery into 70% for machine learning model training and 30% for model testing. The validation cohort was evaluated independently. The data shown here are the gene transcript counts from the training data samples. Please note that the records have been updated with raw sequencing data on Mar 27, 2024.