Global architecture of metabolite distributions across species and its formation mechanisms

Living organisms produce metabolites of many types via their metabolisms. Especially, flavonoids, a kind of secondary metabolites, of plant species are interesting examples. Since plant species are believed to have specific flavonoids with respect to diverse environment, elucidation of design principles of metabolite distributions across plant species is important to understand metabolite diversity and plant evolution. In the previous work, we found heterogeneous connectivity in metabolite distributions, and proposed a simple model to explain a possible origin of heterogeneous connectivity. In this paper, we show further structural properties in the metabolite distribution among families inspired by analogy with plant-animal mutualistic networks: nested structure and modular structure. An earlier model represents that these structural properties in bipartite relationships are determined based on traits of elements and external factors. However, we find that the architecture of metabolite distributions is described by simple evolution processes without trait-based mechanisms by comparison between our model and the earlier model. Our model can better predict nested structure and modular structure in addition to heterogeneous connectivity both qualitatively and quantitatively. This finding implies an alternative possible origin of these structural properties, and suggests simpler formation mechanisms of metabolite distributions across plant species than expected.

💡 Research Summary

**
The paper investigates how secondary metabolites, specifically flavonoids, are distributed among plant species by representing the species‑flavonoid relationships as bipartite networks. Using a curated dataset from Metabolomics.JP, the authors focus on six large plant families (Fabaceae, Asteraceae, Lamiaceae, Rutaceae, Moraceae, Rosaceae), comprising 4 725 species and 6 846 flavonoids, yielding 14 378 species‑flavonoid pairs.

First, the authors quantify three fundamental structural properties of the networks: (i) heterogeneous degree distribution, (ii) nestedness (N), and (iii) modularity (Q). Degree distributions for both plant nodes and flavonoid nodes follow a power‑law with exponential truncation, indicating a few highly connected “hubs” and many low‑degree nodes. Nestedness, measured with the BINMATNEST algorithm, ranges from 0.5 to 0.9, significantly higher than a null model that preserves degree heterogeneity (p < 0.0001). Modularity, evaluated with the Guimerà‑Amaral optimization, is also markedly higher than the null expectation, and the Pearson correlation between N and Q is low (r = 0.345, p = 0.503), confirming that the two patterns are independent.

To explain the origin of these patterns, the paper compares two generative models. The first is the authors’ evolutionary model, a duplication‑divergence process with two parameters: p (probability that a new species appears at each time step) and q (probability that a flavonoid is inherited unchanged from its ancestor). The model proceeds as follows: (1) a new species is created by copying an existing species; each flavonoid of the parent is retained with probability q, otherwise it is lost; (2) with probability 1 − p a new flavonoid is generated by selecting a random existing species‑flavonoid pair and adding a novel flavonoid to that species. Parameters p and q are estimated directly from empirical totals of species (S), flavonoids (F), and interactions (L) using simple algebraic relations (p = S/(S+F), q = 1 − F/L).

The second model is the Bipartite Cooperation (BC) model, a trait‑based, non‑growth framework originally devised for plant‑animal mutualistic networks. In the BC model, species and metabolites are assigned intrinsic traits (e.g., foraging efficiency, reward quality) and external environmental factors; the number of species, metabolites, and interactions are fixed, and links are generated probabilistically based on the trait values.

Model performance is assessed by (a) how well the generated networks reproduce the observed nestedness and modularity, using Pearson correlation coefficient (CC) and root‑mean‑square error (RMSE), and (b) how accurately the degree distributions P(k_S) (species degree) and P(k_F) (flavonoid degree) match the empirical data, measured with a weighted Kolmogorov‑Smirnov statistic (wKS). The evolutionary model achieves CC = 0.949 (RMSE = 0.0039) for nestedness and CC = 0.770 (RMSE = 0.0282) for modularity, outperforming the BC model (CC = 0.946, RMSE = 0.0200 for N; CC = 0.767, RMSE = 0.0708 for Q). For degree distributions, the evolutionary model yields wKS values on the order of 10⁻⁴, indicating an almost perfect fit, especially for flavonoid degrees where the BC model shows substantial deviation (wKS ≈ 10⁻¹–10⁻²).

These results demonstrate that the simple duplication‑divergence dynamics, without invoking explicit trait‑based preferences or external environmental factors, are sufficient to generate the complex architecture observed in real metabolite‑species networks. The authors argue that the evolution of plant species together with gradual diversification of secondary metabolites can naturally give rise to heterogeneous connectivity, nested subsets of metabolites, and modular clusters of species. Consequently, the formation mechanisms of metabolite distributions across plants appear to be governed more by basic evolutionary processes than by sophisticated trait‑based interactions.

In conclusion, the study (1) confirms that plant‑flavonoid bipartite networks exhibit three intertwined structural signatures—heterogeneous degree, nestedness, and modularity—similar to ecological mutualistic networks; (2) shows that a parsimonious evolutionary model reproduces all three signatures more accurately than the established BC model; and (3) suggests that the diversification of secondary metabolites can be understood as an emergent property of simple speciation and metabolite‑innovation events, offering a new perspective for modeling other bipartite biological systems such as microbe‑metabolite or host‑pathogen networks.