Presentation Information
[C06-04]Identifying the phenotype indicator species in chemical reaction networks
*YongJin Huang1, Takashi Okada1,2, Atsushi Mochizuki1 (1. Kyoto University (Japan), 2. RIKEN (iTHEMS) (Japan))
Keywords:
nonlinear dynamics,single-cell metabolomics,network analysis,chemical reaction networks
Cellular phenotypes are highly diverse, reflecting the complex functional states that individual cells can adopt. Traditionally, these phenotypes have been determined using transcriptomic data, whereas recent advances in single-cell technology have shifted the focus toward higher-resolution classification through metabolite phenotyping. Metabolomics provides a direct and comprehensive reflection of cellular functions, yet the vast number of metabolites in biological systems presents a major challenge for phenotypic classification. Since profiling every chemical species is impractical, a fundamental question arises: which subset of species can sufficiently represent the system’s overall state? To address this challenge, we develop a theoretical framework rooted in topological degree theory and structural analyses of chemical reaction networks. By decomposing the network and applying fixed-point theory to each subnetworks using only the structural information, our method identifies a set of key species whose concentrations uniquely determine those of all others in the network. We further implement an algorithm based on this framework and apply it to biochemical pathway databases. Numerical experiments demonstrate that phenotypic classification can be accurately achieved using only the identified indicator species—and that, under data noise, this approach may even outperform classification based on all species. These findings establish a rigorous foundation for selecting indicator species, offering valuable insights into metabolic phenotyping and biomarker discovery.