Construction And Analysis Of Metabolic Network Of Human Genome

The human body is colonized by a large number of microorganisms, collectively referred as the human microbiome. Various microorganisms along with wide range of interactions among them structure the microbial communities as inherently complex ecosystems across different human body site, such as gut, oral and cutaneous. Most of these microbes are still unculturable and uncharacterized. The human microbiome has been proved to play a key role in human health and disease. Futhermore, dynamic shifts in microbial community structure will change the immunity and metabolic function and cause various diseases, including obesity, inflammatory bowel disease (IBD), crohn’s disease and so on. Ttraditional culture-dependent methods are restricted by the small number of cultured species, and often fail to describe the less abundant species. To address this challenge, metagenomics (also referred as environmental and community genomics), a culture-independent technology has been developed and widely used for microbial community analysis. Especially, unlike the initial capillary sequence-based or PCR-based metagenomic approaches, high-throughput metagenomic approaches based on next generation sequencing (NGS) make metagenomic analysis more sensitive, broader and cheaper, providing critical insights into microbe-host interaction in large scale.Previous studies focused mainly on the’parts list’of the microbiome, a nd overlooked the interactions in the complex communities on system-level. Ho wever, an integrated approach to reconstruct the metabolic network by integrati ng metagenomic data with genome-scale metabolic modeling was introduced re cently. This metagenomic system biology method serves the entire microbiome as a single super-organism and utilizes the computational systems biology and complex network theory, providing comprehensive systems-level understanding o f the microbiome by integrating metagenomic data with genome-scale metabolic modeling. A key challenge of applying metagenomics to microbial community is metabolic network reconstruction from metagenomic data. KEGG, the most widely used metabolic database was taken to reconstruct the reference metaboli c network, and metagenomic reads were mapped onto the KEGG database usin g the MG-RAST server. By comparing ’topology-based’ and ’constraint-based’m etabolic network models, ones most commonly framework for microbial cell m etabolism modeling, we found that’topology-based’ method is more suitable for large-scale metagenomic metabolic network. A global reference metabolic netw ork which has 3768 nodes and 81907 edges was reconstructed with the ’topolo gy-based’method. The topology of complex biological networks may provide v aluable information about the functional capacity and help us to understand the observed phenotypes. A large-scale metabolic network was constructed from th e entire enzymatic genes (KO) in any sample. In this metabolic network, enzy mes are connected with directed edges, and a directed edge from enzyme A to enzyme B indicates that a product metabolite of a reaction catalyzed by enzy me A is a substrate metabolite of a reaction catalyzed by enzyme B. We appli ed this metagenomic system-level method to analyze eight human samples with different oral health. Then, the abundances of enzymatic genes was estimated from the functional data annotated with KEGG database. A system-level metab olic network, which has 1009 nodes and 18698 edges, was reconstructed. We c ompared the abundances of enzymatic genes across all sample and identified 2 93 dental caries related genes, which contains 147 enriched and 146 depleted e nzymatic genes. Several Topological features were calculated. Specifically, the depleted enzymatic genes have a significantly higher clustering coefficient comp ared with other enzymatic genes. Furthermore, we developed an R package, m mnet, to implement community-level metabolic network reconstruction. This pac kage is freely available on the Bioconductor project (http://www.bioconductor.or g/packages/devel/bioc/html/mmnet.html). The package also implements a set of f unctions for automatic analysis pipeline construction including functional annota tion of metagenomic reads, abundance estimation of enzymatic genes, communi ty-level metabolic network reconstruction, and integrated network analysis. The package has substantial potentials in metagenomic studies that focus on identify ing system-level variations of human microbiome associated with disease.Moreover, the structure of complex biological systems of microbial communities reflects not only their diversity or function but also the environments they live in. Reverse ecology, graph-theory-based algorithm was introduced to study the ecology of the human microbiome. It was used to analyze metabolic networks of all species in a microbial community, and predicting the metabolic interface of species and their environment. The crucial step of this method was to predict the interaction between organism and its environment, which was called seed set (referred as the exogenously acquired compound). To validate the seed set method, the seed compounds of Buchnera aphidicola was predicted, is composed of 25 compounds (only 9 hava a confidnece level 1). It includes glucose and mannitol which is consistently in previous research. We defined the metabolic competition index and the metabolic complementarity index to represent the interactions between species. Based on this method, we developed RevEcoR, an R package and shiny web application that implements Reverse Ecology algorithm, which can obtain large-scale ecological insights into species’ecology directly from high-throughput metagenomic data. The open source RevEcoR R package is freely available on github https://github.com/yiluheihei/RevEcoR, and the interactive shiny web application is available from http://yiluheihei.shinyapps.io/RevEcoR. We applied RevEcoR to predict the cooperation among several human oral microbiota species whose co-occurrence pattern was well described. And found that Streptococcus oralis and Streptococcus gordonii had the lowest complementarity index and the highest competition index among all pairs. This indicates that these two species are antagonistic, which is in agreement with previous findings. Subsequently, RevEcoR was used for a large-scale human microbiome dataset containing 116 prevalent species. By comparing the species interactions and co-occurrences that can help us to predict whether species that compete with one another tend to co-occur or to exclude, whereas the complementarity index is not correlated with co-occurrence. Our analysis shows that metagenomic systems biology and reverse ecology approach have great potentials for human microbiome research.