bmotif: a package for motif analyses of bipartite networks

Abstract

221. Bipartite networks are widely-used to represent a diverse range of species interactions, such as 23 pollination, herbivory, parasitism and seed dispersal. The structure of these networks is usually 24 characterised by calculating one or more metrics that capture different aspects of network architecture. 25While these metrics capture useful properties of networks, they only consider structure at the scale of 26 the whole network (the macro-scale) or individual species (the micro-scale). 'Meso-scale' structure 27 between these scales is usually ignored, despite representing ecologically-important interactions. 28Network motifs are a framework for capturing this meso-scale structure and are gaining in popularity. 29However, there is no software available in R, the most popular programming language among 30 ecologists, for conducting motif analyses in bipartite networks. Similarly, no mathematical 31 formalisation of bipartite motifs has been developed. 32 2. Here we introduce bmotif: a package for counting motifs, and species positions within motifs, in 33 bipartite networks. Our code is primarily an R package, but we also provide MATLAB and Python code 34 of the core functionality. The software is based on a mathematical framework where, for the first time, 35we derive formal expressions for motif frequencies and the frequencies with which species occur in 36 different positions within motifs. This framework means that analyses with bmotif are fast, making 37 motif methods compatible with the permutational approaches often used in network studies, such as 38 null model analyses. 39 3. We describe the package and demonstrate how it can be used to conduct ecological analyses, using 40 two examples of plant-pollinator networks. We first use motifs to examine the assembly and 41 disassembly of an Arctic plant-pollinator community, and then use them to compare the roles of native 42 and introduced plant species in an unrestored site in Mauritius. 43 4. bmotif will enable motif analyses of a wide range of bipartite ecological networks, allowing future 44 research to characterise these complex networks without discarding important meso-scale structural 45 detail. 46 47

The non-random assembly of functional motifs in plant-pollinator networks

“…We calculated the frequency of all motifs up to five nodes (17 different motifs in total; see Figure 1) for each empirical network, by using the bmotif package version 2.0.2 (Simmons, Sweering, et al, 2019). We focused exclusively on motifs up to five nodes in our analyses for computational reasons and for the ease of visualization and interpretation of a reduced set of less complex network motifs.…”

Motifs in bipartite ecological networks: uncovering indirect interactions

“…At the meso-scale, network structure was described using the frequencies of motifs containing between three and six species, c i , as described above. Motif frequencies were calculated using the 'bmotif ' R package (Simmons et al 2018). We excluded the two-species motif representing a direct interaction between two species because the focus here is on indirect interactions.…”

“…At the meso-scale, species roles were characterised with the vectors describing the frequency with which species occur in all unique positions across motifs containing between three and six species. Species motif roles were calculated using the 'bmotif ' R package (Simmons et al 2018).…”

Community-level reorganizations following migratory pollinator dynamics along a latitudinal gradient

“…Additionally, we compared the signatures of S. rufus to those of two other abundant species, Hylocharis leucotis in Mexico and Calypte anna in California. Motif role signatures were calculated using package bmotif [27]. Motif frequencies were normalized by dividing the position counts for each node by the total number of times that node appears in any of the positions [27].…”

“…Motif role signatures were calculated using package bmotif [27]. Motif frequencies were normalized by dividing the position counts for each node by the total number of times that node appears in any of the positions [27]. Statistical comparisons were done using permutational multivariate analysis of variance (PERMANOVA) using Bray-Curtis as the dissimilarity distance.…”