Abstract
Microbial network construction and analysis is an important tool in microbial ecology. Such networks are often constructed from statistically inferred associations and may not represent ecological interactions. Hence, microbial association networks are error prone and do not necessarily reflect true community structure. We have developed anuran, a toolbox for investigation of noisy networks with null models. Such models allow researchers to generate data under the null hypothesis that all associations are random, supporting identification of nonrandom patterns in groups of association networks. This toolbox compares multiple networks to identify conserved subsets (core association networks, CANs) and other network properties that are shared across all networks. We apply anuran to a time series of fecal samples from 20 women to demonstrate the existence of CANs in a subset of the sampled individuals. Moreover, we use data from the Global Sponge Project to demonstrate that orders of sponges have a larger CAN than expected at random. In conclusion, this toolbox is a resource for investigators wanting to compare microbial networks across conditions, time series, gradients, or hosts.
Highlights
A biologically interesting pattern needs to differ from patterns observed by chance or from patterns generated by processes that are not of interest to the investigator [1]. Such differences can often only be observed by generating data under the sets of rules specified by null models
Null models support the existence of a small core association networks (CANs) in the human gut We inferred 20 networks from time series of stool samples with the fastLSA network inference method [28], one network per person
Relevant patterns through the use of null models and reported that a low-prevalence CAN exists, with associations found in 20–25% of individuals
Summary
A biologically interesting pattern needs to differ from patterns observed by chance or from patterns generated by processes that are not of interest to the investigator [1]. Sample numbers are often insufficient to infer associations accurately In this context, null models can help identify network properties that are different from what is expected based on the null hypothesis that networks are mostly random. We illustrate the power of this strategy by identifying nonrandom CANs in human gut and sponge microbial networks and show that the CANs represent biologically relevant group-specific associations These strategies have been implemented in a software toolbox that evaluates the significance of network or network property comparisons. Set of sets are identified by a combination of intersection numbers: the set of sets 6→10 refers to the models for identification of nonrandom patterns in association networks) difference of intersection 6 and intersection 10 and contains no that generates random networks and assesses properties of these edges present in at least 10 networks. Size This mathematical representation is not implemented directly in the software, as the software takes the set of all edges present in at least four networks and ignores network identity
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
