MaREA: Metabolic feature extraction, enrichment and visualization of RNAseq data

Abstract

The characterization of the metabolic deregulations that distinguish cancer phenotypes, and which might be effectively targeted by ad-hoc therapeutic strategies, is a key open challenge. To this end, we here introduce MaREA (Metabolic Reaction Enrichment Analysis), a computational pipeline that processes cross-sectional RNAseq data to identify the metabolic reactions that are significantly up-/ down-regulated in different sample subgroups. MaREA relies on the definition of a Reaction Activity Score, computed as a function of the expression level of genes encoding for reaction enzymes, which can also be used as an effective metrics to cluster samples into distinct metabolic subgroups. MaREA finally allows to visualize the results in a graphical form directly on metabolic maps. We apply MaREA to distinct cancer datasets and we show that it can produce useful information and new experimental hypotheses on metabolic deregulation of cancer cells, also allowing to stratify patients in metabolic clusters with significantly different survival expectancy.

Integration of single-cell RNA-seq data into metabolic models to characterize tumour cell populations

“…Given the importance of reactive oxygen species (ROS) metabolism observed in [6], we also inserted ROS production and removal pathways. As the original version of the model does not include information on GPRs, such rules have been extracted from Recon 2.2 [36] and included in the HMRcore model, and recently manually curated [16]). We decided to disregard the GPR associated to the complexes I to IV of the electron transport chain in scFBA computations, because it unrealistically requires up to 81 genes (AND rule) and we were not able to accurately tune the rule for these elaborate complexes in [16].…”

“…As the original version of the model does not include information on GPRs, such rules have been extracted from Recon 2.2 [36] and included in the HMRcore model, and recently manually curated [16]). We decided to disregard the GPR associated to the complexes I to IV of the electron transport chain in scFBA computations, because it unrealistically requires up to 81 genes (AND rule) and we were not able to accurately tune the rule for these elaborate complexes in [16]. However the flux trough complexes I to IV should be modulated by the constraints on the last step of the chain (ATP synthase).…”