B overexpression in advanced breast carcinoma and breast cancer cell lines. Consistent with upregulation of EZH2 in aggressive cancers, the clustering revealed high expression of EZH2 in cluster 1, which contained samples from highly aggressive blood and lung cancers. Because HDM and HMT displayed cancer-specific gene expression signature, we asked whether correlations in their target gene expression or in the expression of genes marked with histone methylation are lost in cancer samples. We acquired expression data for KDM5A targets and EZH2 targets from the GSK dataset and applied z-score analysis to study the significance of their combined up- or downregulation in different cancer cell lines samples. We found 
that similar to Digitoxin normal cells, EZH2 targets are strongly biased towards similar expression with “9226994 the H3K27me3 module, while KDM5A module displayed anticorrelative behavior to the expression of EZH2 module and H3K27me3 module . Samples with the highest z-score values for KDM5A module and with the lowest z-score values for EZH2 module were found in leukemia samples of cluster 1, which was consistent with high expression of these modules in normal blood cell types. However, in contrast to normal tissues, there was no tendency for the H3K4me3 module to be overexpressed in the same cancer samples as the KDM5A module . Even more striking differences were displayed in cluster 2, where most of HDMs and HMTs were expressed lower. In cluster 2, there were no significant correlations in expression of any of the four modules. Therefore, cancer-specific HDM/HMT gene expression signature is likely to dictate global changes in histone methylation affecting target gene expression. As the GSK samples from the same tumor types clustered together on the basis of either HDMs or HMTs gene expression, we asked whether we could reveal any cases of HDM and/or HMT co-regulation in human cancer. We computed pair-wise associations among HDMs and HMTs, allowing the identification and visualization of “enrichment”of linked pairs. We have shown a high correlation between KDM5A and the MLL1 gene expression, which may account for the general correlation in expression of KDM5A and its targets. However, in cancer cells we were unable to detect correlation between KDM5A and MLL1 expression, even when ” only hematopoietic cell lines were considered. The highest PCC values, including the anti-correlation between August 2011 | Volume 6 | Issue 8 | e24023 Co-Regulation of Histone-Modifying Enzymes 6 August 2011 | Volume 6 | Issue 8 | e24023 Co-Regulation of Histone-Modifying Enzymes PKDM10B and MLL1, and correlation in KDM5B ASH1 and KDM2AMLL1 pairs, were revealed in colon cancer cell lines. We found a negative correlation of EZH2 and MLL1 expression in colon cancer and positive correlation of EZH2 and KDM1A expression in lung cancer, suggesting cooperation between H3K27 methylation and H3K4 demethylation. In many cases, the revealed associations were clustered by protein families. For example, expression of PKDM10A, -B and -C displayed a lung cancer-specific correlation between each other and with other HDMs and HMTs. Therefore, the new HDM/HMT gene expression signature may explain the lack of correlation between the KDM5A module and H3K4 module in cancer compared to normal tissue. The coordinately regulated HMTs and HDMs may provide functionality in highly specific, yet-to-be-identified, biological processes. When combined with the power of unsupervised hierarchical cl