Purpose:The purpose of this research is to represent the complex system of the rumen microbiome through a multi-layer network, to explore the main biological functions related to methane metabolism. Methods: A three-layer heterogeneous network has been constructed based on rumen metabolites, rumen microbial genes and rumen microbial communities. Node association is calculated based on linear and non-linear association p-value voting combined with the permutation test. This research proposed a method to generate key hubs based on the topological properties of nodes in a multilayer network. Results: 1) A total of 59 individuals (including metabolites, microbial genera, and microbial genes) were found in the topological hubs. Among them, 23 individuals appeared in more than one topologically ranked hub. 2) The metabolite with the highest topological centrality is methane, which is the top one in the 9 topological property rankings. Among the top 10 topologically central 35 microbial genes, based on the reconstruction of the metabolic network, the most involved metabolic pathway is methane metabolism which included 14 microbial genes. 3) There are 3 genera of microbes (including Aeropyrum, Desulfurococcus and Thermosphaera) from the same family and are ranked top 10 in at least 2 topological characteristics. In the topological hub of the Bottleneck, we found that from Aeropyrum and Desulfurococcus of the same family, their indirect paths are associated with methane. And SufD (K09015), a protein related to the family in the same hub, was also found to be positively associated with methane.
Conclusion:The nodes with important topological properties at the metabolite level and the microbial level are associated with methane functions, indicating the consistency of the biological functions of the topological properties between the layers. This method provided a solution for extracting related hubs of highly central nodes and shows potential in the understanding of biological functions. The topological property-Bottleneck showed the most potential by identified hub nodes were found as exhibiting the consistent sulfur reduction functions throughout multilayer networks.