438239 Identification of Cancer Key Metabolic Patterns Using RNA-Seq Data
Identification Of Cancer Key Metabolic Patterns Using RNA-Seq Data
Cankut Cubuk, Marta R. Hidalgo, Jose Carbonell-Caballero and Joaquín Dopazo
Computational Genomics Department, Centro de Investigación Príncipe Felipe (CIPF), Valencia, Spain.
Metabolic abnormalities are the main cofounders of the cancer cells. Characterization of the metabolic biomarkers and targets are of paramount importance for efficient cancer diagnosis and treatment.
In our study, we used K-shortest Elementary Flux Modes (EFMs) , metabolic network topology of Recon1  and the shortest path approach. Since, EFMs guarantee the steady-state conditions, the reaction stoichiometry were used only to generate the reaction-reaction interaction network. In this network, the nodes represent the reactions and the edges represent the connections between the metabolic reactions. Moreover, each of the EFMs was thought as a particular cellular system and exchange reactions were considered as entrance points of the metabolic flow. The shortest paths between exchange reactions were calculated and gene expression data were mapped to their metabolic reactions using gene-protein-reaction (GPR) associations.
Two different behaviours for metabolic reactions were observed between cancer and normal cells; all reaction values inside a shortest path were altered in a similar portion or some particular reactions were altered independently from the others. Based on these observations, we classified the alteration status of the reactions as moderate and disturbed. Later on, we compared the recurrence percentage of these reactions among all shortest paths. This method has been applied to 3 different RNA-Seq datasets [kidney renal clear cell carcinoma (KIRC), breast invasive carcinoma (BRCA), bladder urothelial carcinoma (BLCA)] from The Cancer Genome Atlas (TCGA, https://tcga-data.nci.nih.gov/tcga/).
Unlike the moderately altered reactions, the recurrency percentage of the disturbed reaction series were non-uniform between different cancer types. All these disturbed reactions were tested using SVM classifier and the cancer type prediction accuracy of the our model was %100.
These findings let us to think that the disturbed reactions, profile of their metabolic enzymes and metabolites can be cancer cell specific. The moderately altered reactions were related with lipid metabolism, cell growth and proliferation which are essential cellular functions to differentiate tumor cells.
So far, our results were highly correlated with the literature. Moreover, we obtainded metabolic patterns to distinguish the different cancer types which can be used for cancer type spesific treatment and drug development. However, the missing GPRs are the main limitation of our study. We believe that the novel models such as ReconX series will provide more accurate results. References
1. Rezola, A., Pey, J., de Figueiredo, L., Podhorski, A., Schuster, S., Rubio, A. and Planes, F. (2013) Selection of human tissue-specific elementary flux modes using gene expression data.Bioinformatics, 29, 2009-2016.
2. Duarte, N., Becker, S., Jamshidi, N., Thiele, I., Mo, M., Vo, T., Srivas, R. and Palsson, B. (2007) Global reconstruction of the human metabolic network based on genomic and bibliomic data.Proceedings of the National Academy of Sciences, 104, 1777-1782.