Fig. 4. Numerous BCG-induced antibody features are associated with reduced Mtb burden.
a,c, Spearman correlations between base-10 log(Mtb burden) at necropsy and each plasma (a) and BAL (c) antibody measurement after vaccination. Labeled features above the red dashed line have an adjusted P value (q value) < 0.05 by the Benjamini–Hochberg procedure. b,d, Spearman correlations between base-10 log(Mtb burden) at necropsy and select plasma (b) and BAL (d) antibody measurements. Fold change antibody measurements were subjected to a z-score transformation before plotting. The q values were computed by the Benjamini–Hochberg procedure. e,f, PLS-DA model fit using the antibody features selected by LASSO regularization. e, Graph of the first two latent variables (LVs) of the model. Protected macaques (black) had an Mtb burden < 1,000 colony-forming units at necropsy. Susceptible macaques (gray) had an Mtb burden > 1,000 colony-forming units at necropsy. Ellipses show the 95% confidence intervals. Two-tailed Mann–Whitney U test, P < 2.2 × 10−16. f, Variable importance in the projection (VIP) coefficients on LV1 for each feature.