Graph theory analysis of a human body metabolic network: A systematic and organ-specific study

Abstract

Purposes: Positron emission tomography (PET) imaging is widely used to detect focal lesions or diseases and to study metabolic abnormalities between organs. However, analyzing organ correlations alone does not fully capture the characteristics of the metabolic network. Our work proposes a graph-based analysis method for quantifying the topological properties of the network, both globally and at the nodal level, to detect systemic or single-organ metabolic abnormalities caused by diseases such as lung cancer.

Methods: We used whole-body ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) standardized uptake value (SUV) images from 32 lung cancer patients and 20 healthy controls to construct two-organ glucose metabolism correlation networks at the population level. We calculated five global measures and three nodal centralities for these networks to explore the small-world, rich-club and modular organization in the metabolic network. Additionally, we analyzed the preference for connections significantly affected by lung cancer by dividing organs according to system level and spatial location.

Results: In lung cancer patients, functional segregation in metabolic networks increased (increased $C_p$ , $E_{\mathrm{loc}}$ , and $Q$ , t < 0), whereas functional integration decreased (increased $L_p$ , t < 0, and decreased $E_{\mathrm{glob}}$ , t > 0), indicating more localized and dispersed metabolic activities. At the nodal level, certain organs, such as the pancreas, liver, heart, and right kidney, were no longer hubs in lung cancer patients (decreased nodal centralities, t > 0), whereas the left adrenal gland, left kidney, and left lung showed significantly increased centralities (increased nodal centralities, t < 0). This change suggests compensatory effects between organs. Connections between the nervous and urinary systems, as well as between the upper and middle organs, were more strongly affected by lung cancer (p < 0.05).

Conclusions: Our study demonstrates the utility of graph theory in analyzing PET imaging data to uncover metabolic network abnormalities. We identified significant topological changes and shifts in nodal roles in lung cancer patients, indicating a shift toward localized and segregated metabolic activities. These findings emphasize the need to consider systemic interactions and specific organ connections affected by disease. The impact on connections between the nervous and urinary systems and between the upper and middle regions underscores the modular nature of organ interactions, offering insights into disease mechanisms and potential therapeutic targets.

Keywords: PET imaging; graph theory; lung cancer; metabolic correlation network.