Identifying flow modules in ecological networks using Infomap

Abstract

1. Ecological networks often have a modular structure with groups of nodes that interact strongly within the group and weakly with other nodes in the network. A modular structure can reveal underlying dynamic ecological or evolutionary processes, influence dynamics that operate on the network, and affect network stability. Consequently, detecting modular structures is a fundamental analysis in network ecology.2. Although many ecological networks describe flows, such as biomass flows in food webs, disease transmission or temporal interaction networks, most modularity analyses have ignored these dynamics, which can hinder our understanding of the interplay between structure and dynamics. Here we present Infomap, an established community-detection method that maps flows on networks into modules based on random walks. To illustrate and facilitate the usage of Infomap on ecological networks, we also provide a fully documented repository with technical explanations and examples, and an open-source R package, infomapecology.3. Infomap is a flexible tool that works on many network types, including directed and undirected networks, bipartite (e.g., plant-pollinator) and unipartite (e.g., food webs) networks, and multilayer networks (e.g., spatial, temporal or multiplex networks with several interaction types). Infomap can also use additional relevant ecological data, which we illustrate by investigating how taxonomic affiliation as node metadata influences the modular structure. 4. Flow-based modularity analysis is relevant across all of ecology and transcends to other biological and non-biological disciplines. It includes natural ways to analyse directed networks, take advantage of metadata, and analyse multilayer networks. A dynamical approach to detecting modular structure has strong potential to provide new insights into the organisation of ecological networks, that are more relevant to the system/question at hand, compared to methods that ignore dynamics.

Multilayer diffusion networks as a tool to assess the structure and functioning of fine grain sub-specific plant-pollinator networks

“…For that reason, we used Infomap and its random walk approach (R package "infomapecology", (Farage et al, 2021)) to unveil the module organization of our multilayers. Specifically, Infomap uses the map equation to measure the minimum amount of bits (or code length, L) that are needed to describe the movement of a random walker in and between the modules of a given network partition M (Farage et al, 2021;Rosvall et al, 2009). Then, the algorithm finds the partition that requires the least amount of information to describe modular flows.…”

“…This analogy allows us to study in detail the structure of interactions at increasing scales: 1) micro-scale, by analyzing node roles through properties such as their degree centrality or strength (Gómez & Perfectti, 2012;Gómez, 2019); 2) meso-scale, through the study of local interaction patterns, or motifs (Delmas et al, 2018;Milo et al, 2002;Simmons et al, 2018); and 3) macro-scale, i.e. structural properties of the overall community from a flow-based perspective (Farage et al, 2021;Rosvall et al, 2009). Integrating different scales in a common formalism allows us to explore the relative importance of scale for individual plant reproduction.…”

Multilayer Networks of Plasmid Genetic Similarity Reveal Potential Pathways of Gene Transmission

“…We detected modules using Infomap ---an algorithm based on the movement of a random walker on the network. Infomap is designed specifically for multilayer networks and also measures the amount of flow (based on random-walk movements) contained within each node (the total flow across state nodes in the network sums to 1) [54][55][56]. Flow measurement is particularly suitable for our purposes because it is directly related to the idea of gene exchange [56].…”

The network structure and eco-evolutionary dynamics of CRISPR-induced immune diversification

“…Community detection To find 'communities', or as commonly referred to, 'modules', we used the map equation objective function to calculate the optimal partition of the network 45,46 with the R package infomapecology (version 0.1.2) 47 . Briefly, the map equation is a flow-based and information-theoretic method (implemented with Infomap), which calculates network partitioning based on the movement of a random walker on the network (see for details).…”

“…We shuffled binary matrices (host-spacer and infection networks) with function r00 in package vegan (version 2.5-4) in R), an algorithm that maintains the density of the network. This was done internally using infomapecology (version 0.1.2) 47 . We shuffled weighted matrices (immunity networks in simulated and empirical data) by randomly distributing the interactions (function r00 samp in package vegan (version 2.5-4) in R).…”