A Artificial Intelligence Model (AI) Designed by scientists has created one of the most detailed maps of the brain of the mice to date with more than 1,300 regions and sub -regions -some cartographed for the first time -, A tool that will surely open new neuroscientific exploration channels.
The AI model, developed by scientists from the University of California, San Francisco (UCSF) and the Allen Institute, in Seattle, United States, has rebuilt a brain map that, instead of cell types, represents the different regions and subregions of the brain, some of them unknown so far.
In addition, unlike the previous ones, the new brain map has been made from large -scale spatial transcriptomic data, that is, it has been built entirely with data, rather than with human interpretation.
With this, the map offers a “unprecedented” level of detail that advances in the understanding of the brain by allowing scientists to link specific functions, behaviors and disease states with smaller and more precise cell regions, “which provides a roadmap for new hypotheses and experiments on the functions that these areas perform,” says the authors of the study.
“It’s like moving from a map that only shows continents and countries to one that shows states and cities,” illustrates Bosiljka Tasic, a doctor in molecular genetics of the Allen Institute and one of the authors of the study.
“This new and detailed brain plot, based solely on data and not on human experts, reveals subregions of the mouse brain so far unknown. And, according to decades of neuroscience, the new regions correspond to specialized brain functions still to be discovered.”
The findings have been published on Tuesday in Nature Communications.
A powerful AI model
After this advance is Celltransformer, a powerful artificial intelligence model that can automatically identify important brain subregions from enormous sets of space transcriptomic data.
Space transcriptomic reveals where certain types of cells are found in the brain, but does not reveal the regions of the brain based on its composition.
Celltransformer will allow scientists to define the regions and subdivisions of the brain based on calculations of shared cellular neighborhoods, very similar to how the limits of a city are outlined based on the types of buildings in it.
“Our model is based on the same powerful technology as AI tools such as Chatgpt. Both are based on a” transformative “framework that stands out for its ability to understand the context,” says Abbasi-SL, a professor of neurology and bioengineering at the UCSF and main author of the study.
“While the transformers are often applied to analyze the relationship between the words of a phrase, we use Celltransformer to analyze the relationship between the cells that are close in space. Learn to predict the molecular characteristics of a cell based on its local environment, which allows you to create a detailed map of the general organization of the fabric,” says the researcher.
In addition, the model successfully reproduces known regions of the brain, such as the hippocampus, but, also reveals more detailed subregions and not previously classified in little -known brain regions, such as the reticular core of the midbrain, which plays a complex role in the beginning and release of the movement, the study highlights.