Frequency of host genes appearing in triples. Sets of gene triples were included when the first canonical correlation was at least 0.85 and the second canonical correlation was at least 0.65. These levels were chosen arbitrarily to represent strong multivariate structure as identified by CCA. Genes were ranked by their prevalence of top performing triples. This provided a qualitative profile to select genes that empirically show the strongest potential for being related to the virulence characteristics of the microbiome. (a,b) Genes related to immunity and defense far outperformed the other functional categories. For example, the best two performing intestinal biology genes were in fact also co-listed as immunity and defense genes. (c) In contrast, randomly selected genes did not display any strong multivariate structure with respect to the virulence characteristics of the microbiome.
Schwartz et al. Genome Biology 2012 13:r32 doi:10.1186/gb-2012-13-4-r32