Figure 4.

Top immune biomarker candidates for diagnosis of non-tuberculous mycobacteria (NTM) infection status in cystic fibrosis cohorts. (A) Flow cytometric biomarkers for NTM disease state from cytokine secretion profile data (Wilcoxon rank test). Graph shows log odds ratio for each biomarker candidate for diagnosis of NTM disease state (negative versus exposed state) with 95% CI. (P, probability; FDR, false discover rate; AUC, area under the curve). (B) Stepwise regression model (forward selection) for NTM infection status. Fitted model contains 18 variables shown in table with a final AIC of −30 and an AUC 1. Plots show contribution of each variable to model and fitted value for each patient/control (green dots) when model is applied showing an AUC = 1.