Network structure of the human microbiome

Background/Question/Methods

There are 10 times more non-human cells in the human body than human cells, 100 trillion microbial cells in the human gut, and 100,000 million cells per mL in the human colon, the highest density of microbes recorded in any habitat on the planet. While links have been identified between the composition of the human microbiome and obesity, inflammatory bowel disease, cancer, sexually transmitted diseases, allergies and asthma, and cardiovascular disease among others, very little is known about how the species on and in our bodies actually interact. Network approaches, which have been widely applied to natural ecosystems, represent a promising approach to describing the complex interactions that exist within organisms. We assembled consumer-resource and facilitative interaction networks for human-associated microbial species and source-specific nutrients for five regions of the human body: eye, skin, oral cavity, respiratory tract, and the gastrointestinal tract in order to compare the topological structure of the networks and their robustness to node removal.

Results/Conclusions

The networks were composed of 496 bacteria, 2 archaea (Methanobrevibacter and Methanosphaera), and 2 fungal genera (Malassezia and Candida) along with 43 source-specific nutrients. The number of nodes ranged from 6 for the esophagus to 51 for the respiratory tract. Across all site-specific networks facilitative links outnumbered consumer-resource links by 68% and the complexity, defined as connectance, of facilitative networks was on average 38% higher than the complexity of the consumer-resource networks. The most complex consumer-resource network was for the oral cavity with a connectance of 11% and the least complex consumer-resource network was for the respiratory tract and the large intestine (5%).  The eye also had the most complex facilitative network with a connectance of 83%. The least complex facilitative network was observed for the oral cavity with a connectance of 30%.  Human microbiome consumer-resource interaction networks showed similar node richness but lower connectance than is typically observed in natural ecosystems, although the values were still within the ranges observed in entire ecosystems. The human microbiome networks also show much shorter chain lengths and lower mean trophic level due to the primarily bacterial nature of the communities. Our research represents the first attempt to assemble microbiome networks for the human body.