Summary: A new study reveals how AI-driven deep learning models can decipher the genetic regulatory switches that define brain cell types across species. By analyzing human, mouse and chicken brains, researchers have shown that while some brain cell types have been highly conserved for 320 million years, other brain cell types have evolved uniquely. I discovered that.
This regulatory code not only sheds light on brain evolution, but also provides new tools to study gene regulation in health and disease. The findings highlight how AI can identify different conserved genetic instructions that control brain function between species.
This study also influences understanding of neurological disorders by linking genetic variants to cognitive properties. Researchers are currently expanding their models to study the brains of various animals and human diseases such as Parkinson’s disease.
Important facts
Evolutionary Conservation: Some brain cell regulatory codes have not changed for over 320 million years. AI-powered discovery: Deep learning models can help decipher genetic regulation across species.
Source: Vib
In a new study published in Science, a Belgian research team explores how genetic switches that regulate gene activity define brain cell types across species.
They trained deep learning models on human, mouse, and chicken brain data, and after millions of years of evolution, they are highly conserved between birds and mammals, but other cells evolved differently. I discovered that it was causing the
The findings do not only shed new light on brain evolution. It also provides a powerful tool to study how gene regulation forms different cell types in species and different disease states.
Our brains, and even our whole body, are made up of many different types of cells. They share the same DNA, but these cell types all have their own shapes and functions.
What makes each cell type different is a complex puzzle that researchers have been trying to put together for decades from short DNA sequences acting like switches, stating which genes are on or off. Control it.
The finely tuned regulation of these switches ensures that each type of brain cell performs its unique role using correct genetic instructions from the genome. Scientists call the unique patterns of these genetic switches the regulatory codes.
AI to crack code
Professor Stein Aerts, Vib.ai and Vib-Ku Leuven Center for Brain & Disease Research Study: The basic principles of this regulatory code and how it affects diseases such as cancer and brain damage.
They develop deep learning methods that will help you understand the vast amount of information about gene regulation that is collected from thousands and thousands of individual cells.
“The deep learning models that operate using DNA sequencing codes have been extremely helpful in identifying regulatory mechanisms across different cell types,” explains Aerts.
“Now we wanted to investigate whether this regulatory code could let us know how these cell types are conserved in species.”
One example of where such questions are highly relevant is in the brain. Despite shared developmental trajectories, mammalian and bird brains exhibit significantly different neuroanatomy.
Aerts and his team applied a deep learning model to assess whether existing differences and similarities were reflected in shared or different regulatory codes.
Tools for studying evolution
Nikolai Hecker and Niklas Kempynck are students at Aerts Lab’s Postdoc and PhD, respectively, characterizing and comparing different types of cells across human, mouse and chicken brains, covering approximately 320 million years of evolution We have developed and implemented machine learning models to do so.
However, before they could really compare, they first had to better understand the cell type composition of chicken brains, so they created a comprehensive transcriptome atlas.
“Our research shows how deep learning can be used to characterize and compare different cell types based on regulatory codes,” explains Hecker. “These codes can be used to compare genomes of different species, identify which regulatory codes are evolutionarily conserved, and provide insight into how cell types evolved. .”
The team found that while some regulated cell type codes are highly conserved between birds and mammals, others have evolved in different ways. In particular, the regulatory codes of specific bird neurons are similar to those of deep neurons in the mammalian neocortex.
“Looking directly at the regulatory code offers a great advantage,” adds Kempynck.
Tools for studying disease
This regulatory information is useful not only for understanding evolution. In previous studies, Aerts and his team have already tested that the regulatory code for melanoma (skin cancer) cell state is conserved between mammals and zebrafish. They also identified genome variants in melanoma patients.
The models presented in the current study of brain cell types provide useful tools for studying the association between the effects of genomic variants and their psychological or cognitive properties and disorders.
AERTS: “Ultimately, models that learn genomic regulatory codes have the potential to screen the genome and investigate the presence or absence of a particular cell type or cellular state in any species. This is to study diseases. , it will be a powerful tool to better understand it.”
To the zoo
Aerts and his team have already applied the model in both ways, he says:
“In collaboration with the Zoo Science and Wildlife Rescue Center, we are currently expanding evolutionary modeling to the brains of more animals. At the same time, these AI models will unravel genetic variations associated with Parkinson’s disease. We’ll explore how it can help you.
About this genetics, AI, and evolutionary neuroscience research news
Author: Gunnar de Winter
Source: Vib
Contact: Gunnar de Winter – Vib
Image: Image credited to Neuroscience News
Original Research: Closed Access.
“Comparison of enhancer-driven cell types reveals similarities between mammalian and bird pallium,” Stein Aerts et al. Science
Abstract
Comparison of enhancer-driven cell types reveals similarities between mammalian and bird pallium
introduction
Cell type identity is governed by gene regulatory networks containing cell type-specific combinations of transcription factors (TFS) that bind to genomic enhancer regions.
Placement of the TF binding site forms a cell type-specific enhancer code. Deep learning models trained with single-cell data provide a means to model and characterize enhancer codes at nucleotide resolution. Enhancer codes at such resolutions have yet to characterize the telencephalon brain brain features of mammals, which make up most of the forebrain containing pallium.
Pallium shows significant neuroanatomical differences between mammalian and non-mammalian vertebrates. Most notably, mammalian pallium contains a six-layer neocortex that is absent in all non-mammalian vertebrates, such as birds.
The homology between mammalian and bird pallium has been the subject of decades of debate. Currently, it is unclear whether enhancer codes are conserved throughout vertebrate brains and whether they are beneficial to resolve homologous relationships between species at the cell type level.
basis
To characterize and compare enhancer codes of brain cell types between mammalian pallium and bird pallium, we generate spatially resolved transcriptome data in single-cell multiohms (Scmultiomes) and chicken hindbrain. did.
We compared telencephalon cell types transcriptomes between humans, mice and chickens as a baseline to map cell type similarities between species. We then used variable chromatin accessibility as a proxy to identify potential genomic enhancer regions and assess the specificity of their cell types.
We then trained sequence-based deep learning models in these regions to infer cell type-specific enhancer codes in human, mouse, and chicken neurobrain. We implemented three metrics comparing cell types between species using enhancer codes.
result
Excitatory neurons in the chicken cerebellum are clearly localized to the neuroanatomical regions of pariah, including mesopalium, entopalium, hyperpallium, and nidopallium.
Based on comparisons of transcriptome and enhancer codes, non-neuritis and γ-aminobutyric acid-mediated (GABAergic) cell types are reflected by the combination of conserved TFs of these cell types, and birds and mammals. Shows a high degree of similarity between.
On the other hand, enhancer codes of excitatory neurons in mammalian and avian pallium exhibit a higher degree of divergence. These agreements are only partially in line with existing evolutionary models of homology between vertebrate palelial cell types based on developmental trajectories and brain circuits.
We found that deep excitatory neurons in mammals are most similar to mesopalial neurons, mammalian supernatant layer, piriform cortex, and amygdala neurons.
In vivo enhancer reporter assays were performed to validate predictive responses to mammalian and bird cell types. We show that chicken enhancer sequences show activity in the corresponding mammalian terminal cell type when assayed in mouse brains.
Conclusion
Our study shows that enhancer codes can be exploited to infer cell type correspondences between species along the transcriptome comparison.
A joint comparison of transcriptomes and deep learning-based enhancer codes reveals both predicted and unexpected correspondences between cell types in mammalian and avian telencephalon, resulting from a common amniotic fluid ancestor. It shows conserved regulatory programs that are likely to be co-sourced or diversified.
The proposed enhancer code-based approach is generally applicable and can be used to characterize and compare cell types across species using genomic regulatory codes.
