Having demonstrated that plasma levels of HO-1 and TIMP-4, and to a lesser extent ANC, contribute to the ability to discriminate TB with T2DM from TB without T2DM, we next used PCA models with different inputs to better define the role of these markers in the context of several other biomarkers. In the first PCA model (Fig 5A) in which only HO-1, TIMP-4, and ANC were used, we verified that patients from the two different groups clustered separately but with notable intersection. A second model included data from measurements of several other parameters (Fig 5B) and showed that the two clinical groups clustered separately with no intersection (Fig 5A). To verify the contribution of HO-1, TIMP-4, and neutrophil count in the context of all these other biomarkers, we designed a third model excluding just these three variables. With this modification, the intersection between the groups was dramatically increased (Fig 5A), implying that HO-1, TIMP-4, and ANC are indeed the critical components discriminating TB with T2DM from TB without T2DM in this larger panel. These results were reinforced by cluster analysis using the same combination of the markers from the PCA (Fig 5C). ROC curve analysis ultimately compared the overall performance of the diverse combination of candidate biomarkers and also combined potential confounding factors in distinguishing TB with T2DM from TB without T2DM (Fig 5D). Interestingly, when only information on age, sex, BMI, total cholesterol, HDL, LDL, and triglycerides were considered, the discriminatory performance was poor (area under the curve, 0.58; 95% CI, 0.46-0.70; P = .197) (Fig 5D). The best performance was achieved only when HO-1, TIMP-4, and ANC were considered in the context of several other candidate biomarkers. This approach confirmed that among several markers of inflammation, tissue remodeling, immune activation, and oxidative stress, HO-1, TIMP-4, and ANC are optimal for discrimination of TB with T2DM from TB without T2DM.