Back to the main page.

In the paper, we presented two enrichment experiments to quantitatively evaluate the biological relevance of the discovered subspace differential coexpression patterns: (i) enrichment with ten known cancer-related signaling pathways [1] (ii) enrichment with the 5452 gene sets in the Molecular Signature database (MSigDB) [2]. In addition, (iii) we also did a third enrichment experiment with human GO Ontology annotations [3].

The following figure shows the details of the enrichment results with respect to the above three experiments (Click the figure to view the original size picture).

Note that, as discussed in the paper, due to the limited knowledge about differentially coexpressed patterns, the current stage of differential coexpression pattern mining is still hypothesis generation rather than hypothesis verification, as discussed in Kostka and Spang.20 Indeed, since all the 95 patterns are statistically significant in the permutation test, and all the genes contained in the 95 patterns are known cancer-related genes15 (as described in section 3.1), they can be considered as hypotheses that may lead to new understanding of the interactions among them, and of the relationship between differential coexpression and cancer mechanism. Therefore, in addition to the above standard enrichment analyses, we illustrated and discussed in the paper, several interesting patterns that are enriched with known cancer pathways, or target sets of cancer-related microRNAs and transcription factors.

[1] M. E. Higgins, M. Claremont, J. E. Major, C. Sander, and A. E. Lash. Nucl. Acids Res., 35(suppl 1):D721–726, 2007. http://cbio.mskcc.org/CancerGenes/Select.action

[2] A. Subramanian, P. Tamayo, V. Mootha, S. Mukherjee, B. Ebert, M. Gillette, A. Paulovich, S. Pomeroy, T. Golub, E. Lander, et al. PNAS, 102(43):15545–15550, 2005. Specifically, MSigDB contains 386 positional gene sets, 1892 curated gene sets, 837 motif gene sets, 883 computational gene sets, and 1454 GO gene sets. http://www.broadinstitute.org/gsea/msigdb/

[3] M. Ashburner, et al. Nature genetics, 25(1):25–29, 2000. The GO terms considered had only 2 to 100 human genes annotated to them, since our patterns are generally small.