Fig. 4: The effects of GutCP’s predicted interactions on the gut microbiome and metabolome.
From: Ecology-guided prediction of cross-feeding interactions in the human gut microbiome
a–d Each panel compares the levels of microbial species (a and c; blue) or metabolites (b and d; orange) predicted by our ecological consumer-resource model (x axis) with the experimentally measured levels (y axis); the closer a point is to the marked line (indicating an exactly correct prediction), the better our predictive power. The Pearson correlation coefficients for panels (a) through (d) are as follows: (a) correlation 0.88, P < 10−6, (b) 0.75, P < 10−3, (c) 0.88, P < 10−6, and (d) 0.77, P < 10−6. All P values are estimated using the two-sided t test. The predictions using the original, known-set cross-feeding interactions (production and consumption links) are on the left, and using the additional cross-feeding interactions predicted by GutCP are on the right. e Box plot showing the improvement in prediction error of each metabolite in the fecal metabolome (n = 41 independent samples). In all boxplots, the middle line is the median, the lower and upper hinges correspond to the first and third quartiles, the upper whisker ranges from the hinge to the value 1.5 × IQR (where IQR is the interquartile range) above the hinge and the lower whisker extends from the hinge to the value 1.5 × IQR below the hinge, while all data points failing beyond the range of whiskers are plotted individually. Predictions errors using the original cross-feeding network are in blue, and those with added interactions predicted by GutCP are in red. Central bars indicate median, boxes and whiskers indicate quartiles. Metabolites for which GutCP improved predictions highly are shown in solid bold colors for illustration; those with faded colors represent modest improvements. The shaded gray part of the plot shows new metabolites whose levels GutCP helped predict, but the original cross-feeding network could not.
