BEACON Researchers at Work: Teaching Computational Evolution

This week’s BEACON Researchers at Work post is by University of Texas at Austin postdoc Art Covert.

My name is Art Covert and I’m a teacher. Specifically, I teach the “Computational Evolution Stream” in the Wilke Lab as part of the Freshman Research Initiative (FRI) at the University of Texas at Austin. The FRI is an innovative program at UT Austin that places freshmen interested in research into “streams” which over two semesters train students to do research. Over the two semesters, students develop their own research projects under the guidance of a post-doc level instructor. Students are able to participate in research, in actual research labs, for which they get course-credit, and research groups get many helping hands.

Computational Evolution is a new stream started last year with BEACON support, and is taught by yours truly. In my stream we focus on using the evolution of digital organisms as a proxy for evolution in organic study systems. Digital evolution captures the essence of the evolutionary process with self-replicating computer programs that undergo mutation and selection in a simulated world with limited resources. Researching evolution with computational instantiations lends itself to teaching the scientific method because it’s easy to do controlled experiments, with high statistical power, that will directly answer a well-reasoned hypothesis. Students in my stream not only see evolution occurring before their eyes, but also acquire valuable skills such as doing large-scale data analysis with the python scripting language and learning to use the Sun Grid Engine, the software that powers most of the worlds super-computers, to do their experiments. I use digital-evolution research to engage my students and in the process teach them a variety of skills that they can take with them to almost any research group in undergrad or grad school, as well as industry. Now that we are nearing the end of the first full year of running this stream, I want to share with you three of the projects that my students have been working on.

Box-and-whiskers plots

Figure 1: Two experiments from Terrel's work with paired treatment of an island model at the same migration rate but with different populations structures. On the top, populations of 10,000 digital organisms are divided into 200 subpopulations of size 50. In the bottom plot, populations of equal total size are divided into 25 subpopulations of size 400. We see that as population size of individual sub-populations increases, the optimal migration rate decreases.

Terrel Roane is an undergrad in my stream who has taken over a project I started at Michigan State, examining how genetic drift and migration combine in island models to help populations reach higher fitness than they would achieve otherwise. In island models, populations are subdivided into relatively small subpopulations and are connected by the migration of single organisms, which carry genetic variance between the otherwise isolated groups. Terrel has spent the summer and fall working to understand exactly how these structured populations improve fitness. He has found that when total population size is held constant, larger subpopulations can achieve higher fitness at lower migration rates then smaller subpopulations (Figure 1). Terrel’s results suggest that drift is not the only important factor driving the evolution of island models, and that populations in nature which are structured in similar ways may be able to take advantage of island models with a broader range of migration rates than previously thought.

Jared Carlson-Stevermer has been working in my stream to understand what role deleterious mutations may play in temporally changing fitness landscapes. Digital organisms receive additional bursts of energy if they perform certain logical tasks; therefore, we can manipulate the organism’s fitness landscape by changing which tasks are rewarded. Jared evolves digital organisms which do one task extremely well, then places the best of those organisms in new environments which reward the organisms for doing different tasks (Figure 2, below). By using some clever tests, we can selectively prevent deleterious mutations from entering the evolving populations and determine how important deleterious mutations are to adaptation in changing environments. Jared has found that deleterious mutations tend to play a greater role in environments that are more complex, i.e., those that have several tasks rewarded rather than just one or two. The implication is that deleterious mutations play little or no role when adapting to single-peaked landscapes, but become increasingly more important when adapting to more complex changes in the fitness landscapes. In other words, the more drastic the environmental change, the more important neutral drift may be.

Three fitness curves

Figure 2

Figure 2 (right): Jared tests how changing the environment affects the evolution of digital organism and gives us information on the shape of the fitness landscape. The top plot illustrates the environment all of Jared’s organisms start out in, a single-peaked environment with little or no epistasis, fitness is derived primarily by being able to do task A. We then place a second task in the environment, B or C (middle and bottom plots), which can change the environment in one of two ways. Some new environments correspond to the middle landscape, where peaks A and Bare neutral plateaus to peak AB, and some environments are more like the bottom landscape where peaks A and C are separated by a small deleterious valley which must be traversed to reach peak AC. Jared’s work shows adding more tasks to the environment increases the chance of having a landscape like the bottom plot and increases the role of deleterious mutations, which are needed to cross fitness valleys.

Photograph of Lane Smith and Art Covert

Figure 3: A disheveled post-doc helps Lane optimize the code for his research. FRI students work directly with post-doc level researchers and grad students in the lab.

Lane Smith is a student who is studying the effects of sign-epistasis in sexual organisms. Epistasis is a term used to describe mutations with have different effects on different genetic backgrounds. In their most extreme form, sign-epistatic mutations have the opposite effect when they are combined, individually deleterious mutations can become beneficial. Lane’s project entails tracking all the sign-epistatic interactions in a sexual population between the original ancestor and the final dominant (the most abundant genotype at the end of an experiment). Lane’s project requires that he reconstruct the genealogy over the entire experiment and test every deleterious mutation and their progeny for sign-epistatic interactions, a process that entails looking at millions of genotypes. So far Lane has been able to identify expressed sign-epistatic interactions in several short experiments (with tens-of-thousands of genotypes)

and is working to scale up his experiments to larger runs, and to examine how sign-epistasis is impacted when the pivotal mutational combinations are disrupted by increasing linkage-disequilibrium (Figure 3).

Each of the projects I have described above have a strong potential for publication. We plan to send several of them to the Alife XIII conference. The projects also offer many obvious avenues for future research for the next batch of students to work on. The work being done in this stream has been made possible in large part by our BEACON support, which has allowed us to pool resources from FRI and UT to start a new and exciting research stream that fits squarely within the research and educational goals of both BEACON and FRI. In our next BEACON proposal we plan to ask for funds so that our summer fellows can do digital research at BEACON partner sites during the summer and bring those projects back to UT to complete in the Wilke lab over the fall. It is our hope that students who participate in research at other sites will help to foster collaboration between labs at UT and other BEACON institutions.

For more information, you can contact Art at covertar at gmail dot com.

This entry was posted in BEACON Researchers at Work and tagged , , , . Bookmark the permalink.

Comments are closed.