We will address our hypothesis using a case-control study design. We plan to leverage the existing biorepository of longitudinally collected serum samples through the REMEMBER (Reducing Early Mortality & Early Morbidity by Empiric Tuberculosis Treatment Regimens) ACTG5274 study, which involves serial serum collection from advanced HIV-infected subjects (CD4<50), who did (experimental subjects) and did not develop TB (negative controls) within 24 weeks post-ART. Altogether, serum samples have been collected from approximately 50 cases at the time of TB diagnosis. Serum will be evaluated for the metabolite/miRNA biosignature in cases and negative controls matched on the same two strata used at entry into the parent trial (CD4<25 cells/ml vs. CD4= 26-50 cells/ml and presence vs. absence of poor prognostic factors). Metabolites will be analyzed by LC-MS/MS at Metabolon, and miRNAs will be analyzed by microarrays and reverse-transcriptase PCR (RT-PCR) at Johns Hopkins.