- This event has passed.
Large-scale microbial ecological interaction networks from shotgun metagenomics with the NESRA algorithm
November 22, 2017 | 4:00 pm
Shotgun Metagenomics studies the microbial diversity of specific environments, including the human body. It allows obtaining snapshots of the taxonomic composition and functional potential of a microbial community (microbiome). These microbiome characteristics are niche-specific and are shaped by biochemical factors, such as oxygen and nutrients availability, pH, and temperature. The thousands of bacteria, archaea, viruses, and micro eukaryotes in a microbiome are cross-interacting through ecological relations of mutualism, commensalism, and parasitism. However, despite their relevance, the network of interactions between microbes in microbiomes are not yet well understood, and methods that can scale up to thousands of samples are not available. Here we propose a new method to build microbe-microbe interaction networks, and apply it to thousands of microbiomes from the curatedMetagenomicData resource, a collaboration between CUNY and the University of Trento providing the largest collection of consistently processed shotgun metagenomics datasets available. We considered as features the relative abundances of each species in >4400 human gut microbiome samples from healthy individuals. Preliminary experiments show that our method can easily scale to the number of samples present in the curatedMetagenomicData. Other methods we tested, i.e. CCREPE and SPIEC-EASI, fail to scale to the same number of input samples. The results we obtained provided preliminary insights about the interactions between microbes in the healthy human gut microbiome. In particular, several Bacteroides genera belonging to the Bacteroidetes phylum showed a positive correlation among each other, suggesting that specific host and environment have a similar impact on these phylogenetically related organisms. Interestingly, we found nodes with a high number of connections to involve genera that are usually of low abundance and understudied, thus confirming that some interactions would need to be further explored. We foresee the possibility to use the reconstructed networks to characterize microbiome types.
About the Speaker
Francesco Asnicar is a Ph.D. candidate at the University of Trento, Italy, supervised by Prof. Enrico Blanzieri at the Department of Information Engineering and Computer Science (DISI) and Prof. Nicola Segata at the Laboratory Computational Metagenomics at the Centre for Integrative Biology (CIBIO). Francesco’s main work is studying the microbiome through shotgun metagenomics, by developing new computational analysis tools, with a particular interest on microbial ecology and phylogenomics analysis.