Skip to main content
Fig. 3 | BMC Medical Genomics

Fig. 3

From: Comprehensive discovery of subsample gene expression components by information explanation: therapeutic implications in cancer

Fig. 3

Gene Ontology, KEGG, and protein interaction enrichment for CorEx gene groups. a A circos plot indicates which latent factor groups of the 200 inferred by CorEx are enriched for genes on KEGG pathways (yellow), Gene Ontology terms (orange) and Protein-Protein interactions (blue). A tick mark of the corresponding color is drawn for groups exhibiting enrichment greater than a threshold. Further, the blue PPI tick marks have heights proportional to the relative enrichment (observed/expected). The group results appear in order of decreasing Total Correlation, with the number of genes in a group indicated by the green shading on the outer ring of the circos plot. (darker means a greater number of genes with a range from 4 to 1135) Cutoffs were initially chosen according to the presence of coherent biological signals, however randomization tests indicate corresponding FDR values of 1%, 5%, and 10% for GO, KEGG, and PPI, respectively. The vast majority of CorEx groups show annotation enrichment, typically for all three types. A sample of string network connections are shown on the outside, showing both a diversity of network function and relatively few apparently spuriously associated genes within the groups. b Annotation enrichments of CorEx groups compared to standard clustering methods including hierarchical clustering, k-means, and pca. Enrichment results are based upon 200 clusters for all methods

Back to article page