evolutionary biologist, statistician, nice guy
I am broadly interested in biology and statistics. To keep my interests somewhat focused, I try to select projects that I think are of general importance to evolutionary biology but are also where I believe my joint training in statistics and microbiology is uniquely valuable. I tend to be a bit obsessed with modeling, reproducibility and inferential scope, so much of my work has used laboratory populations that evolve or are manipulated under controlled conditions. Using this approach, I have investigated the nature of epistatic interactions between adaptive mutations, as well as the underlying frequency of such mutations. However, I am also am interested in inference problems involving observational data, and currently am working on projects which investigate evolutionary dynamics and macro-evolutionary processes for wild populations. Below are some links to pages that describe some of my work, feel free to browse and contact me if you have any questions. I always love to chat or correspond about science.
Evolution is mutations causing phenotypes and the stochastic dynamics that follow. Central to understanding this process is understanding the epistatic interactions that can occur between different mutations. These interactions have been theorized to lead to the evolution of sex, ploidy, dominance and reproductive isolation. It has also been shown that they can greatly reduce the number of adaptive trajectories available to evolution. If we are to understand evolution then, we must understand epistasis.
By analyzing all beneficial mutations that occur within a protein I helped show that their are typical patterns of epistasis within proteins and that these patterns have dramatic consequences for evolutionary dynamics, this research is described here.
However, we might expect that interactions between mutations that are not within the same protein but rather lie in different proteins, metabolic networks or biological modules behave differently, as the functional relationship is quite different. I also investigated patterns of epistatic interactions between proteins and show that they have their own predominant pattern, quite different from the intra-protein pattern, but equally consequential. This is described here.
Both of these studies involved micro-dissections of single adaptive walks. To scale up this understanding and integrate it in a broader physiological framework, I have built a high-throughput robotic system for phenotypic measurements and experimental evolution described here. Currently, I am using this system in combination with a novel population marking technique to generate a broad sampling of adaptive potential in the aerobic microbe Methylobacterium extorquens.
To build models that translate laboratory results to the biological world, one needs to have at least an approximate idea of the dynamics of that world. Understanding how fast mutations appear, how strong selection is and how frequent recombination is are all critical to translating patterns we see in the lab.
I am fortunate enough to have access to data for a well defined population of a pathogenic microbe that killed about 225 million birds in the US while simultaneously being heavily sampled and studied, which allows for strong insights into evolutionary dynamics in the wild. You can read more about this work here.
Often times though, we have a very interesting biological question but a very poor or inexact dataset with which to answer it. This is really what led to the growth of statistics and I think some of the hardest problems in biology today are in the comparative methods field. Unfortunately though, many of the methods used in that field are not statistically rigorous and have shortcomings involving the method of inference or the inductive bias. I am investigating ways to correct these issues as well as assessing what if any effect they have on our conclusions. More can be read about this here.
I am also interested in how what I learn about epistatic interaction in the lab says about the way we build models for association studies that try to guess genetic effects from existing phenotypes. You can read some of my musings on this here.