PS 31-176 - Can causal inference help microbial ecologists ask better questions?

Abstract: Microbes play important functional roles that affect the health of plants, animals, and ecosystems. Microbial ecologists are interested in how these microbial functions respond to variation in microbial community composition, especially when microbial community composition changes in response to some external driver. This implies the following causal chain: "driver" - > "microbes" - > "function". There is an emerging field of inquiry, called "causal inference," that seeks to understand the implications and effect sizes of these kinds of causal structures that can be expressed in the form of directed acyclic graphs (DAGs). The goal of my presentation is to introduce microbial ecologists to the basic toolkit of causal inference so that they can begin to create and analyze DAGs to help them better understand their systems. I will show that the causal framing of "driver" - > "microbes" - > "function" represents a fundamental problem for microbial ecology, because information about "microbes" blocks the causal linkage between "driver" and "function." In these cases, it will almost always be more efficient to just study the "driver" and the "function," and ignore the microbial community altogether. This also partially explains why so many microbial ecology studies end up only describing the "driver" - > "microbes" part of the chain. I propose some alternative framings that can lead to more profitable microbial ecology research. For example, we can usefully apply microbial ecology to cases where microbes partially mediate the effect of the "driver" on the "function." We can also use DAGs to identify specific microbial taxa that both respond to the "driver" and directly affect the "function." My hope is that increased use of DAGs and other tools of causal inference can help microbial ecologist design more sophisticated studies and lead to more conceptual and theoretical advances in the field.