Visualization of Microbial Dynamics
Modern microbiome studies often revolve around the dynamics of and differences between microbial communities. For example, what is the behavior of the microbiome after a treatment of antibiotics, or do microbial signatures differ between treatment and control groups. Interpretation of these complex data have inspired various statistical methods, but the potential for informative data visualization in this domain has remained relatively unexplored. In Treelapse (like timelapses, but for trees), we adapt methods developed in the data visualization research literature in order to study microbial dynamics.
The data used in these visualization experiments come from the study,
DiGiulio, Daniel B., et al. "Temporal and spatial variation of the human microbiota during pregnancy." Proceedings of the National Academy of Sciences 112.35 (2015): 11060-11065.which can be obtained here. In the DOI tree, Timeboxes, and Treeboxes views give the abundances for the vaginal site in subject 10101. Node sizes and edge widths represent abundances averaged across all timepoints. In the DOI Sankey, we average samples for each of the CSTs identified in this study. In the sparklines example, we use data from subject 10032. In the future, we plan on making these techniques applicable to generic data sets through an an htmlwidgets interface to phyloseq.
These visual experiments are only preliminary, and this project remains under active development. All code is publically available on github. For questions, please contact Kris Sankaran. More resources for microbiome data analysis can be found on the Holmes lab website. Thanks to Jasmine Sankaran for suggesting the name Treelapse.