Ecological networks (Chair: CHRISTIAN BICK)

Ecological networks model species-to-species interactions, and are intended to be predictive models for an ecosystem. We infer ecological networks from observational data, and often assume precision and recall, i.e., that (1) a network link reflects a true pairwise functional relationship between species, and (2) all true relationship are modelled as links. Unfortunately, for ecosystems such as the soil, where the species are both numerous and microscopic, functional information is rare because the individual interactions cannot be observed in the field. Instead, spatial co-occurrence networks are inferred from sampling data. We ask the question: how accurate are these spatial co-occurrence networks of microorganisms, as inferred with current soil-sampling methods? No prior result on this question is available in soil ecology.Method. An agent-based model with biologically realistic behaviour and parametrisation simulates a plot of land, with true trophic links between species. We observe the spatial co-occurrence that these trophic links naturally produce in space (with or without an equilibrium state, step 2). We also simulate the taking of samples from this spatial distribution of species. Finally, we evaluate the accuracy of the co-occurrence network inferred from samples, against the true co-occurrence of the plot.Results. We find that biological properties other than the interactions, such as species diversity, can be estimated with relatively low error with sample pooling. On the other hand, the inference of the co-occurrence network is poor. We see high errors of the pairwise link weights, with mean errors around 0.5 and very large standard deviations between experiments. The co-occurrence network inferred is thus both inaccurate and unstable (explaining the large differences seen among algorithms for co-occurrence inference), and this is intuitively explainable in spatial terms.

Mobile genetic elements appear in nearly all microbial systems and affect ecological processes and, subsequently, macroscopic quantities such as system stability and composition. However, little is known about how the spread of these elements through horizontal gene transfer is affected by ecological processes, or how this spread affects population dynamics. We discuss a model that allows one to leverage techniques from epidemiology to predict whether mobile genetic elements are maintained. Our results show that maintenance depends on the species abundance distribution, thus advancing our understanding of the connection between ecological and epidemiological processes.

In green plants, there exist two different types of chlorophylls, chlorophyll-a and chlorophyll-b, in light-harvesting proteins. Although the characteristics of individual chlorophyll are well understood, the advantages of their coexistence yet to be understood. In this study, we simulate excitation energy transfer within the entire photosystem II supercomplex by employing network analysis integrated with quantum dynamic calculations. We consider the energy transfer process as a Markov process and numerically trace the energy flow during photosynthesis. The result shows that the natural chlorophyll composition allows the excited energy to preferentially flow through specific domains that act as safety valves, preventing downstream overflow. We also investigate the influence of the proportion between chlorophyll-a and chlorophyll-b on the photosystem by comparing various chlorophyll compositions. Our findings suggest that the light-harvesting proteins in a photosystem II supercomplex achieve evolutionary advantages with the natural chlorophyll-a/b ratio, capturing light energy efficiently and safely across various light intensities. Through this study, we propose a novel method to investigate the photosystem with which one can better understand how green plants harvest light energy and adapt to changing environmental conditions.

Natural ecosystems exhibit remarkable diversity and stability, yet the underlying mechanisms remain unclear. Pathogens, often overlooked in ecological models, play a crucial role in maintaining these dynamics. Focusing on microbial systems—characterised by large pathogen diversity—we address the challenges of scaling models to accommodate large, complex infection networks. Employing a pathogen-centric framework, we investigate how pathogen traits and competition for limited hosts influence maintenance and, consequently, their ecological impact. These findings offer insights into the interplay between ecological and epidemiological processes, providing a foundation for future studies on pathogen-driven stability and community composition.

Mean abundance distributions (MADs) are important observables to understand biodiversity and community structure in ecological systems, often following a lognormal distribution in microbial communities. Using a stochastic Lotka-Volterra model, we investigate the impact of interaction networks and species-specific carrying capacities on MADs and extinction dynamics. Interaction networks are modeled as random Erdős-Rényi graphs, with both directed and undirected types. Directed networks refer to commensalism or amensalism, while undirected networks describe competition between species. For dense networks, MADs deviate from the lognormal distribution, becoming increasingly skewed. Our results show that species abundances depend on carrying capacities, while network connectivity has a lesser influence. Extinction rates rise with network density and interaction strength, also occurring within small sub-communities represented by disconnected network components. In brief, we have studied the effect of the carrying capacity and the interaction network density on species abundances and extinction rates. This work, which is currently in preparation, highlights the influence of the distribution of carrying capacities and the interaction network on community composition.

Graphs have long served as effective models for pairwise interactions, providing insights into fundamental properties of trophic and mutualistic networks. However, it is now recognized that non-pairwise (higher-order) interactions significantly influence the stability of underlying dynamical systems. Hypergraphs, where edges can connect arbitrary subsets of vertices, have been proposed as models capable of incorporating such interactions. Yet, their application to interaction modifications often obscures the roles of individual vertices and amplifies centrality measures.In this work, we present an extension of the hypergraph concept that more consistently represents complex higher-order interactions, including the modification of information. We introduce the concept of a pangraph, as an extension of the ubergraph [2] into directed graphs, which allows edges to start and/or end at other edges at arbitrary levels of nesting. In this talk, we will explore the properties of pangraphs, their relationship to classical directed graphs through the Levi representation, and propose centrality measures for these structures. Additionally, we will discuss potential methods for extending the pangraph concept to metric graphs dynamics.Finally, we demonstrate two compelling applications of pangraphs. The first applies the pangraph framework to structures the theory of Petri nets with catalysts. The second challenges the hypothesis about the importance of interaction modifications presented in [1]. Using a real-world coffee ecosystem database, we show that the results in [1] can be interpreted as the amplification of centrality measures in the hypergraph approach.

Many ecological communities exhibit alternative stable states where the outcome of community assembly depends on both the nature of species interactions and the history of the assembly process. Although only a single stable state can be realized locally, multiple alternative states could be realized at the regional level if local communities undergo sufficiently distinct assembly histories. However, alternative states are not easily maintained because dispersal in the system tends to homogenize local communities, creating correlations that effectively reduce the number of distinct histories, and thus, decreasing the probability of realizing different alternative states. Despite this, empirical studies have found natural systems in which alternative states persist even in the presence of frequent dispersal. To address the apparent gap between theoretical understanding and observations, we conducted numerical experiments of a two-species competition model embedded in a spatially-explicit landscape where alternative states arise when species compete more strongly with heterospecifics than conspecifics. We tested the hypothesis that assortative dispersal – individuals preferentially disperse into patches containing more conspecifics – can expedite the formation of spatial structure that acts to buffer the homogenising effects of dispersal and facilitate the maintenance of alternative states. Contrary to expectations, we found that although assortative dispersal facilitates the maintenance of alternative states it limits the extent of spatial structure that forms. This occurs because assortative dispersal minimizes the impact of heterospecifics in adjacent patches, preventing local dominance from percolating across the whole landscape. Furthermore, we found that these results are robust against variation in the strengths of interspecific interaction, disturbance rates, and dispersal costs. We discuss our results in the context of niche theory.

When studying a complex system it is often useful to think of the system as a network of interacting units. One can then ask if some properties of the entire network are already rooted in a small part of the network–a network motif.We say a motif is functional when observing one copy of a motif in a large network already guarantees a network-level property, regardless of the rest of the network.A famous example of a functional ecological motif is the exploitative competition in food webs, where the presence of two species competing for a shared resource precludes the existence of a stable equilibrium for the whole network. In ecology the discovery of this motif has led to the formulation of the competitive-exclusion principle that has shaped the field and inspired many subsequent advances. However, other examples of motifs with such direct implications on ecological stability are not known.In this talk we explain why the exploitative competition is a functional stability motif and also why such motifs are rare in food webs and other networked systems, such as epidemic spreading and supply chain dynamics. Building on these results we then discuss more broadly under which conditions functional motifs exist. Perhaps more importantly we show that functional motifs that have implications other than asymptotic instability are common in networks from applications.