Fig. 3: Experimental and clinical data supporting in silico predictions.

A Area under the curve (AUC) measurements for fungal growth in presence of selected metabolites in a series of concentration dilution experiments. AUCs were determined for three replicates. Mean change compared to medium-only controls is shown with standard deviations (SD) as error bars. B Bacterial abundance and growth rates were obtained using MetaPhlAn2 and GRiD 1.2, respectively. Modeled species were arranged according to the Open Tree of Life 10.4 [34]. Annotation rings from inner to outer: Significant correlations between C. albicans abundance and bacterial abundance (magenta stars) or bacterial GRiD score (green stars, Spearman’s coefficient, p < 0.05); species with GRiD score greater than 1 in at least one sample (black triangles); in silico interaction coefficients from GSMM analysis (blue to red); sample bacterial abundance (N = 26, yellow to purple) sorted by C. albicans abundance (highest abundance is outermost ring). C Regression performance for all (left panel) or active (GRiD value > 1 in at least one patient, right panel) species. Each dot is a ratio of inhibitors to promoters for a patient. Values for inhibitors and promoters were calculated by summing products of bacterial abundance x in silico coefficient for C. albicans. D Area under the curve (AUC) of ordinal regression analysis for seven species with significant correlations with C. albicans abundance or GRiD values as shown in B (left panel). AUC of ordinal regression analysis for 29 selected species (see main text). The regression model performance was achieved by using GSMM based metabolic coefficients (Coefficients), bacterial relative abundance (Abundances) or the product of both (Products).