Three faces of metabolites: Pathways, localizations and network positions

TL;DR

This study explores the organization of metabolism by analyzing pathways, localizations, and network clusters, revealing a core-periphery structure with modules and overlaps, and emphasizing the importance of multiple classification schemes.

Contribution

It introduces a topological analysis of metabolite categories using social-network techniques, providing new insights into metabolic organization and module identification.

Findings

Metabolism has a core-periphery structure with dense modules and localized organelles.

Pathways connect multiple network clusters and localizations laterally.

Modules are relatively independent, overlapping more than chance, but not perfectly aligned.

Abstract

To understand the system-wide organization of metabolism, different lines of study have devised different categorizations of metabolites. The relationship and difference between categories can provide new insights for a more detailed description of the organization of metabolism. In this study, we investigate the relative organization of three categorizations of metabolites -- pathways, subcellular localizations and network clusters, by block-model techniques borrowed from social-network studies and further characterize the categories from topological point of view. The picture of the metabolism we obtain is that of peripheral modules, characterized both by being dense network clusters and localized to organelles, connected by a central, highly connected core. Pathways typically run through several network clusters and localizations, connecting them laterally. The strong overlap between organelles and network clusters suggest that these are natural "modules" -- relatively independent sub-systems. The different categorizations divide the core metabolism differently suggesting that this, if possible, should be not be treated as a module on par with the organelles. Although overlapping more than chance, none of the pathways correspond very closely to a network cluster or localization. This, we believe, highlights the benefits of different orthogonal classifications and future experimental categorizations based on simple principles.