AI reveals the evolution of genetic control elements in the mammalian cerebellum

Artificial intelligence allows us to track the evolution of genetic control elements in the developing mammalian cerebellum. An international research team led by biologists from Heidelberg University, the Vlaams Institute for Biotechnology, and the University of Leuven in Belgium has developed an advanced AI model that can predict the activity of these elements based solely on their DNA sequence. Using these models, scientists were also able to track evolutionary changes in regulatory programs and identify changes that are unique to the human lineage.

Genetic control elements are DNA sequences that determine when and where genes are switched on. Changes in the activity of these elements can lead to evolutionary innovations such as brain expansion. One brain region that has significantly expanded during human evolution is the cerebellum. Beyond its role in movement and balance, the cerebellum also contributes to cognition, emotion, and language.

Tracing the evolution of gene control elements has long been difficult due to their rapid evolutionary turnover and limited understanding of how their activity is encoded in DNA sequences. ”

Professor Henrik Kessmann, Research Group Leader, Center for Molecular Biology, University of Heidelberg (ZMBH)

To fill this knowledge gap, researchers took advantage of advances in artificial intelligence. “Customized tools for AI-based analysis of comprehensive and complex datasets in the life sciences have made it possible to decipher sequence grammars and thus the genetically encoded activity profiles of these regulatory elements,” said Professor Stein Aerts, a computational biologist at the Vlaams Institute at the University of Leuven, who co-led the research study with Professor Kaesmann.

The researchers used modern sequencing techniques to map the activity of these elements in individual cells in the developing cerebellum of humans, bonobos, macaques, marmosets, mice and possums. Using this unique dataset, we trained a model based on machine learning to be able to predict the activity of regulatory elements directly from each DNA sequence. These AI models were not only able to model the activity of these elements in the six species studied, but also accurately predicted activity across other mammals. “This shows that the sequence rules that define the genetic control elements of cerebellar cell types have been highly conserved throughout mammalian evolution,” explains Dr. Ioannis Salopoulos, a former doctoral student in Kassman’s group and co-lead author of a paper published on the latest findings with Mari Sepp, Ph.D., a postdoctoral fellow in Kassman’s lab, and Tetsuya Yamada, a doctoral candidate.

Based on these findings, the scientists leveraged the ability of AI models to recognize conserved sequence rules to predict the activity of regulatory elements in 240 mammalian species. For each human element, the researchers were able to determine whether the corresponding sequence was also active in other mammals. This allowed us to reconstruct the evolutionary history of human regulatory programs at high resolution and identify programs that may have contributed to important evolutionary innovations in the human cerebellum. For example, new regulatory elements have been discovered near genes. THRBencodes the thyroid hormone receptor found in all vertebrates. This new element allowed the gene to also function in cerebellar stem cells. According to Kessmann, this may have contributed to the evolutionary expansion of the human cerebellum. “The ability to reuse evolutionarily old genes for new functions is an important mechanism by which evolution drives innovation,” emphasizes the Heidelberg molecular biologist.

In addition to the teams in Heidelberg and Leuven, researchers from Göttingen and Leipzig, Hungary and the United Kingdom contributed to the study. The project was funded by various organizations and foundations, including the European Research Council, the European Molecular Biology Agency, and the Simons Foundation. The research results were published in a magazine science.