ABSTRACT
In recent years, there has been strong interest in examining the microbiome and its impact on human health and the environment. By leveraging modern sequencing technologies, investigators can quickly determine the composition of a given microbial sample. At the same time, the same investigations often yield an array of categorical and numerical metadata derived from the sequenced samples such as immunohistochemical measures or locality information. Understanding how the microbiome data is associated with this external metadata is essential in developing targeted treatments for chronic diseases or proposing bacteria-modulated host responses. While many R or Python libraries and command-line tools have been developed for specific analysis purposes, there are still relatively few tools to facilitate open-ended data exploration and hypothesis generation. Here we introduce Mian, an open-source web framework to interactively visualize or run a suite of statistical and feature selection tools on the microbiome to identify important taxonomic groups in the context of any provided categorical or numerical metadata. Visualizations include boxplots, correlation networks, and PCA or NMDS scatterplots. Tools include Fisher’s Exact Test, Boruta feature selection, alpha and beta diversity, and differential and correlational analysis. Mian supports multiple standard representations of the OTU table as input and optionally subsamples the data during the upload process. Users can also filter and aggregate the OTU table at different taxonomic levels and dynamically adjust analysis parameters to see how the visualizations, results, and statistical measures change in real-time. Mian is freely available at: miandata.org