Natural selection and other evolutionary forces leave characteristic signatures on the genetic variation within populations. My group uses a combination of theoretical and experimental approaches to study how this genetic variation is created and maintained, and to develop methods to infer the evolutionary history of populations from the variation observed in sequence data. Our focus is primarily on the dynamics and population genetics of natural selection in asexual populations such as microbes and viruses. We are developing new approaches to population genetic theory to better understand the structure of genetic variation in these populations. To complement this theoretical work, we have developed high-throughput techniques which allow us to directly observe the evolution of thousands of experimental budding yeast populations simultaneously, tracking changes in fitness and other phenotypic characteristics and correlating these with the evolution of genetic variation within and between populations.
Theoretical Population Genetics
One focus of current work in population genetics has been to infer the action of natural selection based on sequence data. This typically relies on an understanding of the expected patterns of genetic variation in the absence of selection pressures, and an intuitive sense of the way in which selection would lead to deviations from these neutral expectations. Yet population genetic theory has struggled to provide precise predictions about what we should expect sequence data to look like in the presence of different types of selective pressures. The problem is that each mutation originally occurs in a genome which has some set of other mutations. In asexual populations or on short distance scales in sexual genomes, selection can only act on these sets of mutations in their entirety. The mutations are physically linked, so their fates are not independent. In these situations, we know how neutrality shouldn’t look, but not how selection should.
A main goal of our theoretical work is to fill this gap in our understanding of the action of selection when linkage is important. Because selection tends to amplify the effects of rare mutations, this often involves situations where fluctuations are crucial, and out-of-equilibrium behavior is important. We have introduced several new approaches to analyzing the evolutionary dynamics in such situations by developing techniques to understand the statistical behavior of large numbers of interacting random processes that are driven by fluctuations in a few rare types. This work has opened up new ways to understand the structure of genealogies and thus the statistics of genetic variation in these populations. We are currently developing various types of effective coalescent approaches which provide a general framework for analyzing sequence data in the presence of selection.
Experimental Evolution
Sequence data from natural populations usually only provides information from a snapshot in time. We can get much more information on evolutionary dynamics by actually observing these dynamics while they occur. Experimental evolution offers the opportunity to do just that, so we are currently using evolution of experimental yeast populations to directly observe what sorts of processes are typically important, and hence to guide our theoretical work. We also use carefully constructed experiments to measure important evolutionary parameters such as distributions of mutational and epistatic effects.
A main challenge of experimental evolution is that much of the important dynamics that determines the long-term fate of particular mutations happens when those mutations are rare and therefore hard to observe experimentally. We have developed a system which allows us to quickly detect certain types of mutations when they are at frequencies of fractions of a percent within a population, allowing us to directly observe these dynamics of certain rare mutants. This class of observable mutations provides a useful probe into evolutionary dynamics which can then be further supplemented by targeted sequencing efforts.
A second challenge is that even though some types of events are very unlikely to happen on laboratory timescales, they could still be crucial for long-term evolution of natural populations. Further, there is an inherent randomness to evolution, so identical populations will often evolve very differently. We have therefore developed high-throughput approaches to experimental evolution which allow us to maintain thousands of evolving lines simultaneously. This has allowed us to observe classes of mutational events which are thought to be quite important in nature but are rare enough that they have not been seen in a systematic way in earlier experiments. It also gives us power to make use of the information contained in the variation we observe in outcomes between identically evolved populations. We are currently taking advantage of these high-throughput techniques to investigate a variety of questions, such as the role of geographic structure in adaptation and the statistical structure of epistasis among beneficial mutations.
Selected Publications
Weissman, D. W., M. M. Desai, D. S. Fisher, and M. W. Feldman (2009). "The Rate at Which Asexual Populations Cross Fitness Valleys." Theoretical Population Biology 75:286-300.
Desai, M. M. and J. B. Plotkin (2008). "Finite Site Effects on the Polymorphism Frequency Spectrum Under Directional Selection." Genetics 180:2175-91.
Desai, M. M. and D. S. Fisher (2007). "Beneficial Mutation-Selection Balance and the Effect of Linkage on Positive Selection." Genetics 176:1759-98.
Desai, M. M., D. S. Fisher, and A. W. Murray (2007). "The Speed of Evolution and Maintenance of Variation in Asexual Populations." Current Biology 17:385-94.
