The two most significantly enriched pathways were o idative phosphorylation and ubiquinone metabolism. selleck compound The p values increase rapidly to 10 5 within the top five ranked pathways. The best performing e pression based signature is enriched for cytoskeleton remodelling, regu lation of cell cycle checkpoint G1 S and regulation of cellular metabolism. The only other very significant enrichments were obtained with the network derived signature based on the betweenness centrality of nodes. The enriched path ways were involved in protein folding and regulation of G1 S transition. Noteworthy enrichments were also found for the signatures based on the degree centrality of nodes in the interaction net work. All three of these signatures yielded several highly enriched path ways for nucleotide metabolism.

The results of all the enrichment calculations are provided as an E cel spreadsheet, see additional file 1. The best performing optimised gene signature with 128 genes showed a similar result as the one obtained for the highest absolute and highest mean e pression signatures o idative phosphorylation and ubiquinone metabolism were consis tently the most significant pathways across several of the optimised signatures from different runs of the genetic algorithm. The low p values for o idative phosphorylation are due to the large size of this pathway compared to all other pathways. This pathway contains several large comple es of the respiratory chain and is composed of a total of 105 proteins. The ubiquinone metabolism path way counts 74 proteins, 46 of which pertain again to mammalian comple 1.

The constituents of the o ida tive phosphorylation pathway, especially so the parts of the electron transport chain composed of comple es 1 to 4, are highly e pressed in the mitochondria of all cells. Furthermore, rapidly dividing cancerous cells in culture, such as the ones used to derive the e pression values used in this study, also require a lot of energy. thus, high e pression levels are to be e pected for mem bers of the respiratory chain. Naturally, such highly e pressed genes were Brefeldin_A selected for inclusion into the sig nature of genes with highest e pression, which in turn e plains the observed enrichment. A more intriguing fact is that the same enrichment is observed for the best performing optimised signature. This suggests that there is at least some overlap in the functionality of the genes of the two signatures given the large size of the mammalian comple es 1 to 4 there need not be an overlap of the same genes, but in genes that belong to the same comple . The optimised signa ture therefore contains a significant part of genes that have a high level of e pression overall, whereas the other genes were selected by the algorithm for other reasons.