A subset of cutaneous melanomas contains high expression of genes related to oxidative phosphorylation determined using consensus clustering

Despite new treatment options, prognosis of many melanoma patients remains poor and new treatment strategies are still needed. Stratification in molecular subclasses may aid therapeutic development. Therefore, we performed consensus clustering of 405 RNA microarray profiles obtained from the Gene Expression Omnibus (GEO) and 469 RNA sequencing profiles obtained from The Cancer Genome Atlas (TCGA), all from tumor samples of patients with cutaneous melanoma of any stage. We biologically annotated each cluster based on the enrichment of pre-defined Hallmark gene sets as assessed by Gene Set Enrichment Analysis (GSEA). To evaluate the concordance between both datasets Spearman's rank correlations between the Z-transformed P-values of matched genes as obtained by pair-wise class comparison were calculated. Consensus clustering revealed four molecular clusters in the GEO dataset. These clusters were biologically annotated and labeled 'immune' (27% of samples), 'estrogen response/p53-pathway' (estrogen/p53) (34% of samples), 'cell cycle' (15% of samples) and 'oxidative phosphorylation' (24% of samples). In the TCGA database we also identified four clusters and annotated them 'immune' (36% of samples), 'estrogen/p530 (7% of samples), 'cell cycle' (24% of samples) and 'oxidative phosphorylation' (32% of samples) according to the GSEA results. Spearman's rank correlations were 0.61 for the 'immune', 0.47 for the 'estrogen/p53', 0.60 for the 'cell cycle' and 0.68 for the oxidative phosporylation clusters. These data indicate four robust molecular subclasses in cutaneous melanoma that may guide future research and development of treatment strategies. Inhibition of mitochondrial function is of interest as a potential treatment strategy for tumors in the oxidative phosphorylation cluster.