FIG 6.

M. tuberculosis-specific antibody profiles distinguish LTBI from ATB irrespective of HIV status. Supervised analyses comparing M. tuberculosis-specific antibody responses elicited by ATB and LTBI individuals. (A and D) Visualization from PLS-DA models trained using the LASSO-selected features in the HIV-negative (A) and HIV-positive (B) subset of individuals. Ellipses represent 95% confidence intervals. CV accuracy of the models and permuted models are indicated. The accuracy of each model is significantly higher than that of the respective permuted model (Mann-Whitney U test HIV negative model, P < 2.2e⁻¹⁶; Mann-Whitney U test HIV positive model, P < 2.2e⁻¹⁶). (B and E) LASSO-selected features from the HIV-negative (B) and HIV-positive (E) models are plotted on VIP plots. Variables with VIP scores greater than 1 contribute most to separation across LV1. LASSO-selected features significantly different across the groups during univariate analyses are indicated. Kruskal-Wallis with Dunn’s multiple-comparison test was used. Adjusted P values are as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. (C and F) Radar plots of each LASSO-selected feature and its significant correlates in the HIV-negative (C) and HIV-positive (F) models. Median Z-score of each group is plotted for a given feature. Significant correlations were defined as those with q values of <0.01. The q values were calculated using the Benjamini-Hochberg procedure (77).