Recently I’ve been working a lot on evident, a Python package that streamlines calculation of effect sizes and power analysis on microbiome diversity data. As I was benchmarking evident for a paper, I found myself trying it out on a couple different diversity metrics. For example, I was interested in seeing if the effect size of diversity differences in sex was different between Unweighted UniFrac and RPCA. This seemed to me an unexplored area of research.
Often, we tend to not put that much thought into our choice of alpha/beta diversity metric. However, this choice can have profound impacts on our conclusions. Differences in community structure that may be observed in one metric may be hidden in other metrics. I wanted to develop a way for researchers to easily benchmark different diversity metrics on their own data to see how each one affects community structures.
Enter xebec. xebec is a Snakemake pipeline built with cookiecutter for diversity difference effect size benchmarking. With xebec, you first point to a feature table, sample metadata file, and phylogenetic tree. Then, you run the Snakemake pipeline which generates the effect size calculations and interactive visualizations showing the effect sizes of each diversity metric on each sample metadata column.
I hope that researchers will use xebec to investigate how the choice of diversity metric can affect downstream results. That way, we can generate more targeted hypotheses and analyses. Below is an example interactive visualization generated by xebec using Bokeh. You can move around, mouse over points, and zoom in. The data used in this study were from this study (Qiita ID: 11402).