ATria: A novel centrality algorithm applied to biological networks

Large-scale biological networks such as gene regulatory and PPI have become commonplace through systems biology. Figure 1 shows a microbial social network, which attempts to infer interactions between microbes within a community from metagenomics studies. In particular this is a co-occurence network, where green/red edges respectively represent positive/negative correlations, or how strongly two bacterial taxa tend to co-infect samples. We may be interested in bacterial taxa that drive community behavior, initial infectors of a community, or the effects of structural changes. Answers to these questions can come from discovering ”important” nodes in these biological networks, which is the goal of social network centrality. Depending on the definition of importance used, many different centrality notions currently exist. We suggest three notions that are potentially important to biological networks, and especially to microbial social networks: 1) For each club or high density subgraph, a dominant or leader node that is responsible for connecting many individuals and driving club behavior. 2) A villain node or common enemy, defined as having many strong negative edges to a club. 3) A bridge node that connects two or more clubs, thus having the ability to link different social circles.