Extraction and Prioritization of a Gene-Cancer-By-Survival Network Involved in Homeostasis of Intracellular Calcium...

Regulation of intracellular calcium concentrations, [Ca is important in maintaining the viability of normal as well as cancer cells and can be mediated by tumor microenvironment. Calcium release-activated calcium channel protein (ORAI) calcium channels on the plasma membrane (PM) become physically connected by stromal interaction molecules (STIMs) to the endoplasmic reticulum (ER), on which paralogous receptors of inositol phosphate and of ryanodine are also present along with ATP2A/SERCA (sarco/endoplasmic reticulum calcium ATPases) subunits (also known as PM-ER geneset). Proper expression of this functionally and physically interconnected geneset is essential for the maintenance of [Ca , yet has not been interrogated as a whole for its role in cancer prognosis using multivariable Cox regression. In the present study, we examined whether the expression profile of the PM-ER geneset exhibited prognostic significance across different cancers found in The Cancer Genome Atlas (TCGA) by generating gene-cancer-by-survival networks, in which the nodes represented either genes or cancers and the edges, the logarithmically transformed hazard ratios for overall survival (OS). We then applied network clustering to identify the gene-cancer subnetworks with high connectivity, among which uveal melanoma (UVM) emerged exhibiting the highest degree of genes (k = 10). BAP1, a well-known [Ca regulator and a tumor suppressor, was not found to be significant in predicting OS by PM-ER geneset for UVM, yet it was for several others, including mesothelioma (MESO). Moreover, the best subset of the PM-ER geneset obtained by lasso predicted OS in the TCGA UVM cohort with an area under the receiver operating characteristics (AUC) of 91.4%, comparable to or better than previous prognostic signatures in the literature. Our findings indicate that homeostasis of [Ca is an essential determinant of prognosis in multiple cancers and particularly in UVM. The proposed gene-cancer-by-survival network approach can be extended with other gene sets as well as different survival types.