As an ecologist studying microbial communities in natural and anthropogenic systems, it’s difficult sometimes not to frame other parts of my life within statistical and ecological theories. I often find myself thinking about social networks—representations of relationships among discrete objects—and how much about our everyday lives can be quantified using network theory.

Take my family circle for example. Growing up, my grandparents hosted many holidays, birthdays, and social events, bringing together aunts, uncles, cousins, siblings, children, and grandchildren. We ate a lot of birthday cake. My grandmothers and grandfathers connected to a large number of other family members and represent keystone nexuses (we would say “high centrality” members) within the family. Without these strong connections, the entire network may become less connected and other, more locally centralized networks (network theorists would call this “disjointed”) may form.

It’s easy to determine who are the keystone connections within a family consisting of perhaps a few dozen members. But how to determine who has the most functional importance in a group of say, 1000? What about 12,000 members?

This is the first challenge we faced when studying microbial communities in a smooth brome invaded grassland in central Saskatchewan. Smooth brome is an invasive grass from Europe that makes for great forage for farm animals, though produces enormous swaths of monocultures where it invades and can have tremendous effects on biodiversity and ecosystem services through changes in nutrient cycling¹ and microbial composition^2,3. 

Unearthing the keystone microbes driving ecological change under plant invasion is difficult because the number of samples in sequencing data are typically much smaller than the number of taxa in each sample. Enter the second challenge, known as the “small n, large p” problem, where you’ve got far more subjects (e.g., amplicon sequence variants) than measurements (e.g., samples), this turns classical statistics sideways as there is little statistical power available to complete community analyses⁴.

Some solutions to this problem in microbial space have included removal of rare taxa under the assumption that they add noise to multivariate analyses. However, there is considerable evidence that some rare taxa can have important ecological roles in complex communities^5,6, and hence removing rare taxa risks misinterpretation of the data⁷.

To meet these challenges in our smooth brome invasion dataset, we developed an iterative data winnowing (think separating the wheat from the chaff) process for inferring community development across environmental gradients. The gist: build a network of all microbial taxa, arrange centralities in decreasing order, identify the sweet spot at which removing taxa decreases treatment effect, and use those taxa for downstream analyses.

What did we find? Analyzing the winnowed microbial community across a gradient of smooth brome invasion, we found that core microbial networks became more connected as smooth brome become more dominant. What this means for invaded grasslands is this: the relationships that form as smooth brome invades are strong and likely represent a new equilibrium for grassland communities. Smooth brome is there to stay.

So much like my family unit, times of stress and change can build strong connections among members. Though I suspect there is much less birthday cake involved when it comes to invaded grasslands.

References

1. Piper, C. L., Lamb, E. G. & Siciliano, S. D. Smooth brome changes gross soil nitrogen cycling processes during invasion of a rough fescue grassland. Plant Ecology 216, 235-246, doi:10.1007/s11258-014-0431-y (2015).
2. Mamet, S. D., Lamb, E. G., Piper, C. L., Winsley, T. & Siciliano, S. D. Archaea and bacteria mediate the effects of native species root loss on fungi during plant invasion. ISME J 11, 1261-1275, doi:10.1038/ismej.2016.205 (2017).
3. Piper, C. L., Siciliano, S. D., Winsley, T. & Lamb, E. G. Smooth brome invasion increases rare soil bacterial species prevalence, bacterial species richness and evenness. Journal of Ecology 103, 386-396, doi:10.1111/1365-2745.12356 (2015).
4. Faust, K. & Raes, J. Microbial interactions: from networks to models. Nature Reviews Microbiology 10, 538, doi:10.1038/nrmicro2832 (2012).
5. Lynch, M. D. J. & Neufeld, J. D. Ecology and exploration of the rare biosphere. Nature Reviews Microbiology 13, 217-229, doi:10.1038/nrmicro3400 (2015).
6. Cao, Y., Larsen, D. P. & Thorne, R. S. Rare species in multivariate analysis for bioassessment: some consideration. Journal of the North American Benthological Society 20, 144–153 (2001).
7. Shade, A. et al. Conditionally Rare Taxa Disproportionately Contribute to Temporal Changes in Microbial Diversity. mBio 5, doi:10.1128/mBio.01371-14 (2014).