Bipartite graphs for metagenomic data analysis and visualization

Abstract

Metagenomics became very popular after expansion of next-generation sequencing techniques that allowed simple implementation of extensive studies. With a target gene sequencing approach, an identification of organisms in a metagenome is quite effortless since only a small reference database of the particular gene is needed. Moreover, by counting the copies of individual genes, also quantitative analysis can be applied. Unfortunately, current bioinformatics tools aim mainly on the analysis of a single metagenome. A cluster analysis, a heatmap of correlation coefficients, biclustering or other statistics techniques can only show relations inside the metagenome or the relation between the metagenome composition and other parameters. On the other hand, there is a lack of tools to provide a comparative analysis of two or more metagenomes. Suitable properties for this kind of analysis can be found in a bipartite graph. Here, we present a novel workflow for finding the suitable representation of metagenomic data to provide a comparative analysis of metagenomes. The resulting graph can take into account information about the actual composition of the metagenome as well as the environment it relates to. Thus, it can provide different view of the data to the naked eye that can complement other techniques such as principal coordinate analysis.