Harnessing the power of machine learning, the researchers built a framework for analyzing the factors that contribute most significantly to a species' genetic diversity.
The study, published recently in the journal Molecular Systematics and Evolution, suggests that different processes have shaped genetic variation in two species of amphibians native to northeastern Brazil: the Brazilian brown frog and the common toad.
Results showed that genetic variation in sibylline toads was primarily shaped by population dynamic events in response to habitat changes that have occurred over the past 100,000 years. In contrast, genetic diversity in granular toads was primarily shaped by modern landscape factors. Toads that are relatively isolated by geographic distance or harsh habitats are more likely to be genetically distinct.
Previous studies have looked at the effects of historical demographic and landscape factors on amphibian genetic diversity, but the datasets on these factors were conducted separately, making it difficult to determine which were most important. Now, rather than manually inferring which was more important, the researchers on this paper are the first to use artificial intelligence to examine how both processes shape genetic diversity equally.
“Before this study, we weren't able to examine both effects in the same framework, so we had to ask them separately,” said Brian Carstens, a professor of evolution, ecology and organismal biology at Ohio State University and co-author of the study. “AI allows us to simulate the processes that are happening ecologically today as well as those that were happening during that ancient evolutionary event, and then compare the results to the actual data we collected from these frogs.”
The vast amount of data that has become available to geneticists and other wildlife biologists over the past few decades can make it difficult for researchers to pinpoint specific factors that may be important in a given experiment, Carstens said. But by integrating large amounts of information into simulations and taking those factors into account in a single analysis, it's possible to get a more complete record of a species' development.
“Building and training an AI model takes a long time, but we wanted a model that could capture the range of potential variation in the history of a species in a way that was as faithful as possible to our knowledge of the biology of the system,” Carstens said.
For example, although the species investigated in this study inhabit the same region, they have many differences in their natural histories: despite both eggs and larvae being fully aquatic, sibylline toads breed continuously in basements throughout the rainy season, whereas granular toad reproduction occurs in bursts as they are dependent on heavy rainfall.
Combined with a machine learning approach, the researchers' simulations resulted in 100% support for their model scenarios for the historical explanation for the spread of the sibylline toad, and over 99% support for the granular toad.
One reason their model is so accurate is that it can take into account recent demographic events, including measuring how events like human development and habitat change have affected animals' genetic diversity over long periods of time.
But even when using AI, researchers need to be careful to avoid misleading patterns in the results, Carstens said.
“The analyses we do can't capture all the factors that have been important to these species over millions of years,” he said, “so we have to consider a range of possibilities without being so broad that basically any model can fit the data.”
Still, advances in technology are enabling researchers to answer niche ecological questions and test new hypotheses, and their work is a precursor to creating an upgraded machine learning framework that can be applied to unique investigations of other species, Carstens said.
“We're going to continue to use these AI tools in different ways and in different combinations to understand evolutionary history,” Carstens says, “and as we continue to learn, the tools we use will change and get better.”
Co-author Emmanuel M. Fonseca, who earned his doctorate from Ohio State University in 2022, is the author of the study. The research was supported by the Ohio Supercomputer Center, the U.S. National Science Foundation, and Brazil's Coordenação de Aperfeiçoamento de Pessoal de Nível Superior.
/Public Release. This material from the originating organization/author may be out of date and has been edited for clarity, style and length. Mirage.News does not take any organizational stance or position and all views, positions and conclusions expressed here are solely those of the authors. Read the full article here.
