Figure 4

First and second canonical correlations between host gene sets and microbial virulence characteristics. Horizontal lines in the density plots are at 0.5, and the vertical lines are at 0.85. These cutoffs were chosen arbitrarily to emphasize enrichment in the upper-right quadrant of the plot that is suggestive of increased multivariate structure as identified by CCA. (a) First and second canonical correlations between triples of immunity and defense genes and virulence variables are shown. There are increased canonical correlations in the upper-right corner of the plot, suggesting an enriched multivariate relationship between the immunity and defense genes and microbiome virulence characteristics as compared to, for example, the set of random genes shown in (d). (b) Intestinal biology genes did not show the same level of enrichment of canonical correlations as the immunity and defense genes. (c) We analyzed 1,000 random sets each containing 660 genes in an analogous manner to the immunity and defense gene analysis (a). Of these, 969 random sets resulted in less than 12% of analyzed gene triples having first canonical correlation >0.85 and second canonical correlation >0.5. (d) An example random gene CCA plot. Additional examples are given in Additional file 6.