Leuven, 14 February 2025 - A Belgian research team, in a study published in Science, used deep learning to explore how genetic switches define brain cell types across species. They trained models on human, mouse, and chicken brain data, discovering that some cell types are highly conserved between birds and mammals after millions of years of evolution, while others evolved differently.
The study sheds light on brain evolution and provides tools for studying how gene regulation shapes different cell types across species or disease states. Brain cells, like all cells in the body, share the same DNA but differ in shape and function. Researchers have been working to understand what makes each cell type unique, focusing on short DNA sequences that act as switches, controlling gene activity. The regulation of these switches ensures each brain cell uses the correct genetic instructions to perform its role. Scientists call the patterns of these genetic switches a regulatory code.
Prof. Stein Aerts and his team at VIB.AI [VIB Artificial Intelligence] and the VIB-KU Leuven Center for Brain & Disease Research study the regulatory code's principles and its impact on diseases like cancer or brain disorders. They develop deep learning methods to analyze gene regulation information from thousands of individual cells.
Aerts explains, "Deep-learning models working with the DNA sequence code have helped us enormously to identify regulatory mechanisms across different cell types. Now, we wanted to explore whether this regulatory code could also inform us on how these cell types are conserved across species."
Mammal and bird brains have different neuroanatomy [nerve structure], despite shared developmental trajectories. Aerts' team applied deep learning models to assess whether these differences and similarities are reflected in shared or divergent regulatory codes.
Nikolai Hecker and Niklas Kempynck, a postdoc [postdoctoral researcher] and Ph.D. student respectively in the Aerts lab, developed machine learning models to compare cell types across human, mouse, and chicken brains, spanning approximately 320 million years of evolution. They first created a transcriptomic atlas [a comprehensive map of all RNA molecules in a cell or organism] to understand the chicken brain's cell type composition.
Hecker states, "Our study demonstrates how we can use deep learning to characterize and compare different cell types based on their regulatory codes. We can use these codes to compare genomes of different species, identify which regulatory codes have been evolutionarily preserved, and gain insights into how cell types have evolved."
The team found that some regulatory cell type codes are highly conserved between birds and mammals, while others have evolved differently. The regulatory codes for certain bird neurons resemble those of deep-layer neurons in the mammalian neocortex [the outer layer of the brain].
Kempynck adds, "Looking directly at the regulatory code presents a significant advantage. It can tell us which regulatory principles are shared across species, even if the DNA sequence itself has changed."
Aerts' team previously verified that regulatory codes for melanoma (skin cancer) cell states are conserved between mammals and zebrafish and identified variants in melanoma patient genomes. The models from the current study on brain cell types provide tools to study the impact of genomic variants and their association with mental or cognitive traits and disorders.
Aerts says, "Ultimately, models that learn the genomic regulatory code hold the potential to screen genomes and investigate the presence or absence of specific cell types or cell states in any species. This would be a powerful tool to study and better understand disease."
Aerts and his team are expanding their evolutionary modeling to more animal brains, from fish to deer, hedgehogs, and capybaras, in collaboration with Zoo Science and Wildlife Rescue Center. They are also exploring how these AI models can help unravel genetic variation linked to Parkinson's disease.