Expression-Based Inference of Cancer Metabolic Flux Differences

Abstract

1Cancer cells display numerous differences in metabolic regulation and flux distribution from noncancerous cells, which are necessary to support increased cancer cell growth. However, current experimental methods cannot accurately measure such metabolic flux differences genome-wide. To address this short-coming, we apply FALCON, a computational algorithm for inferring metabolic fluxes from gene expression data, to analyze data from The Cancer Genome Atlas (TCGA). We found several major differences between tumor and control tissue metabolism. Cancer tissues have a considerably stronger correlation between RNA-seq expression and inferred metabolic flux, which may indicate a more streamlined and efficient use of metabolism. Cancer metabolic fluxes generally have high correlation with their normal control counterparts in the same tissue, but surprisingly, there are several cases where tumor samples in one tissue have even higher correlation with control samples in another tissue. Finally, we found several pathways that frequently have divergent flux between tumor and control samples. Among these are several previously implicated in tumorigenesis, including sphingolipid metabolism, methionine and cysteine synthesis, and bile acid transformations. Together, these findings show how cancer metabolism differs from normal tissues and may be targeted in order to control cancer progression.