Graph embedding is a transformation of nodes of a network into a set of vectors. A good embedding should capture the underlying graph topology and structure, node-to-node relationship, and other relevant information about the graph, its subgraphs, and nodes themselves. If these objectives are achieved, an embedding is a meaningful, understandable, and often compressed representation of a network. Unfortunately, selecting the best embedding is a challenging task and very often requires domain experts. In this paper, we extend the framework for evaluating graph embeddings that was recently introduced in [15]. Now, the framework assigns two scores, local and global, to each embedding that measure the quality of an evaluated embedding for tasks that require good representation of local and, respectively, global properties of the network. The best embedding, if needed, can be selected in an unsupervised way, or the framework can identify a few embeddings that are worth further investigation. The framework is flexible and scalable and can deal with undirected/directed and weighted/unweighted graphs.

In the last years, there were a growing number of studies using the metric H2′ to calculate complementary specialization in host–parasite interaction networks. However, only a few studies have explored the sensitivity of H2′ to network dimensions (i.e. species richness and number of interactions), which consequently could generate studies that are not comparable among them or lead to biased conclusions. In this study, we used the recent published study conducted by Rivera-García et al. in 2016 involving host–bat fly networks as an example to call attention to the risk of using H2′ to calculate specialization for small matrices. After conducting analyses based on both empirical and simulated data, we show that H2′ values are strongly affected by randomly allocation of species interactions to another cell in the matrix for small networks and that therefore the results and conclusions presented in Rivera-García et al. in 2016 are only an artefact of the dataset used. Therefore, we fully recommended taking into account the careful use of small networks to measuring specialization in host–parasite interactions.

Non-random assemblages have been described as a common pattern of flea co-occurrence across mainland host species. However, to date, patterns of flea co-occurrence on islands are unknown. The present work investigates, on one hand, whether the decrease in the number of species on islands affects the pattern of flea co-occurrence, and on the other hand, how the cost of higher flea burdens affects host body mass. The study was carried out in the Canary Islands (Spain) using null models to analyse flea co-occurrence on Rattus rattus and Mus musculus. Results supported aggregation of flea species in Mus but not in Rattus, probably due to the relationship between abundance and both prevalence and intensity of infection of the main flea species parasitizing Mus. In addition, heavy individuals of both rodent species showed the highest flea burdens as well as higher species richness, probably due to the continued accumulation of fleas throughout life and/or immunological resistance mechanisms. Whatever the mechanisms involved, it is clear that co-occurrence and high parasite intensities do not imply a detrimental biological cost for the rodents of the Canary Islands.

Species of Miconia are considered keystone plant resources for frugivorous birds in Neotropical forests, but for other ecosystems little of their ecological role is known. The fruiting phenology and the composition of frugivores of four Miconia species in savanna and palm swamp from the Brazilian Neotropical savanna were studied in three sites from November 2005 to May 2011. The hypothesis tested was that plants from different habitats share their frugivores and consequently promote links between habitats. Through focal plant observations (30–50 h per species in each site), 668 visits by 47 species of birds were recorded and plants from different habitats shared most of the frugivores (49–97%). The fruiting of Miconia chamissois in the palm swamp during the period of fruit scarcity (dry season) was accompanied by an enhancement in the frugivore bird richness and abundance in this habitat, providing indirect evidence of resource tracking. Bird species which primarily dwell in savanna recorded consuming fruits in palm swamps during the resource-scarce season is taken as evidence of landscape supplementation. Miconia assemblage studied here seems to promote a link between two adjacent habitats in the Neotropical savanna from Central Brazil, a link which is likely to be common in this naturally patchy ecosystem.

Properties of ecological networks facilitate the understanding of interaction patterns in host-parasite systems as well as the importance of each species in the interaction structure of a community. The present study evaluates the network structure, functional role of all species and patterns of parasite co-occurrence in a host-parasite network to determine the organization level of a host-parasite system consisting of 170 taxa of gastrointestinal metazoans of 39 marine fish species on the coast of Brazil. The network proved to be nested and modular, with a low degree of connectance. Host-parasite interactions were influenced by host phylogeny. Randomness in parasite co-occurrence was observed in most modules and component communities, although species segregation patterns were also observed. The low degree of connectance in the network may be the cause of properties such as nestedness and modularity, which indicate the presence of a high number of peripheral species. Segregation patterns among parasite species in modules underscore the role of host specificity. Knowledge of ecological networks allows detection of keystone species for the maintenance of biodiversity and the conduction of further studies on the stability of networks in relation to frequent environmental changes.

We tested the hypothesis that species of a diverse leaf-litter ant community are separated by the temperature preferences of their broods along a thermal gradient. Therefore, temperature preferences of brood-tending workers from 41 ant species co-occurring in four types (alluvial, limestone, kerangas and dipterocarp forest) of primary rain forest in Sarawak, Malaysia were measured in an experimental set-up. Preferred temperatures of species ranged from 16.0 °C to 31.7 °C, with the median at 25.8 °C. The ten commonest species (n ≥ 4) showed significantly different temperature preferences. In particular, species of the genus Pheidole differed clearly in their preferences over a broad range of temperatures. Temperature preferences varied significantly among ant assemblages from different forest types and nest sites. Experimentally obtained temperature preferences of species correlated significantly with vegetation density in the plots inhabited by the respective species, but not with plot canopy cover. When we tested the temperature preferences of all ant species with null models for niche overlap, we found a significant niche separation only among the tested species from the kerangas. Our results suggest that nest temperature is an important ecological factor for leaf-litter ants in rain forests on Borneo, but other factors may override its influence during community assembly.

The search for nested subset patterns has become a powerful tool for understanding the processes shaping parasite communities. Here, we re-examine the results of past studies on nestedness in parasite communities, to assess how sensitive they are to the analytical method used. Using the metric N and the null model RANDOM1, the first method available to study nested patterns, early studies concluded that nestedness was infrequent in parasite communities. In contrast later studies, using instead the metric T and the nestedness temperature calculator (NTC), found that nested subset patterns were very common in parasite communities. Recently, a new algorithm, the binary matrix nestedness temperature calculator (BINMATNEST), has been proposed to quantify nestedness. Using data on 31 helminth communities of fish hosts, we show that applying the NTC yields consistently more significant nested patterns than when N and RANDOM1 are used on the same data. The use of BINMATNEST produced results that depend on the choice of the null model. To provide a benchmark, a straightforward comparison between the observed frequencies of co-occurrences of species with those expected from their prevalence under random assembly was also made for each community. This test indicates that random structure occurs in practically all communities, even those where one of the nestedness analyses found a significant pattern. We demonstrate that the probability of finding a nested pattern in a parasite community depends entirely on the metric and null model chosen for analysis.