Computational Methods
Evolutionary and ecological interactions and feedbacks are complex and studying adaptive evolution in response to ecological conditions is challenging. Genomic signals of natural selection can be confounded by demographic changes and other selection pressures, and the methods on the forefront of theoretical development are typically developed for model organisms and can be inappropriate for use with non-model species. For this reason, my work includes the development and careful application of computational and statistical methods that combine theory, big data, and machine learning to study non- and new model species.
Detecting selection with machine learning
|
Simulations for understanding non-model species
|
Population genetic theory in species of conservation concern
|
My machine learning method, Flex-sweep (Lauterbur et al. 2022a, in review) to identify diverse selective sweeps genome-wide that improves on previous methods in its versatility and ability to identify older sweeps.
|
Recently, I led a push to expand the species catalog for the stdpopsim framework for standardizing population genomic simulations for methods development and analysis (Lauterbur et al. 2022b, in review). Originally stdpopsim included six model species; now it includes an additional 15 species, most non- or new models.
|
Coalescent theory is a powerful tool with which to make inferences about the evolutionary history of a population, but it relies on some fundamental assumptions, such as a large effective population size. How much influence does violating that assumption – as in endangered species with very small or declining populations – have on how well we can use this theory to understand what’s going on in endangered populations? (Lauterbur 2019)
In addition, I help to lead the Effective Population Size working group of the Global Bat Network, in which we have begun a meta-analysis of concepts, applications, and methods of calculating Ne, especially as it concerns conservation practitioners. |