| Popis: |
BackgroundBiomarker discovery for pulmonary tuberculosis (TB) may be accelerated by modeling human genotypic diversity and phenotypic responses toMycobacterium tuberculosis(Mtb). To meet these objectives, we use the Diversity Outbred (DO) mouse population and apply novel classifiers to identify informative biomarkers from multidimensional data sets.MethodTo identify biomarkers, we infected DO mice with aerosolizedMtbconfirmed a human-like spectrum of phenotypes, examined gene expression, and inflammatory and immune mediators in the lungs. We measured 11 proteins in 453Mtb-infected and 29 non-infected mice. We have searched all combinations of six classification algorithms and 239 biomarker subsets and independently validated the selected classifiers. Finally, we selected two mouse lung biomarkers to test as candidate biomarkers of active TB, measuring their diagnostic performance in human sera acquired from the Foundation for Innovative New Diagnostics.FindingsDO mice discovered two translationally relevant biomarkers, CXCL1 and MMP8 that accurately diagnosed active TB in humans with > 90% sensitivity and specificity compared to controls. We identified them through the two classifiers that accurately diagnosed supersusceptible DO mice with early-onset TB: Logistic Regression using MMP8 as a single biomarker, and Gradient Tree Boosting using a panel of 4 biomarkers (CXCL1, CXCL2, TNF, IL-10).InterpretationThis work confirms that the DO population models human responses and can accelerate discovery of translationally relevant TB biomarkers.FundingSupport was provided by NIH R21 AI115038; NIH R01 HL145411; NIH UL1-TR001430; and the American Lung Association Biomedical Research Grant RG-349504. |
