This AI-generated representation of a mouse’s encephalon shows how a fresh cerebral blueprint mapped varied sectors of the body part, with some being newly discovered.(Image credit: University of California, San Francisco)
Employing an AI architecture akin to ChatGPT, investigators have formulated a novel map of the murine cerebrum that captures formerly unmapped zones of the framework in striking particulars.
The charting, issued this past Tuesday (Oct. 7) in the scientific publication Nature Communications, delineates 1,300 areas of the brain and constitutes the foremost effort to characterize cerebral zones independently of manual input from individuals. The study’s writers, hailing from the University of California, San Francisco (UCSF) and the Allen Institute for Cell Science, express optimism that this endeavor will empower scientists to create similar tissue charts across the whole physical form.
You may like
-
Blueprint of 600,000 neural cells revises the textbook on decision-making processes in the brain
-
Academics have just engineered a new AI simulating the human brain — it’s surpassing LLMs such as ChatGPT in logical thinking
-
‘Minibrains’ unearth mysteries of how vital cerebral cells emerge in utero
The method also portrays the placements of specific cells within the tissue realm. Prior cell catalogs of the murine cerebrum were built upon this data. Nevertheless, arranging the data obtained from these analyses into an all-encompassing cerebral blueprint presents a considerable hurdle. For prior mappings of the brain, scientists were required to manually specify each segment of the mapping to define particular brain locations and where the documented cells fit within them. The present investigation circumvented this onerous task.
The spatial transcriptomics information employed for the fresh mapping encompassed details on the function of 500 to 1,000 genes in each examined cell. According to study co-author Reza Abbasi-Asl, a professor of neurology and bioengineering at UCSF, this degree of intricacy renders the data dissection demanding. Abbasi-Asl also noted that delineating cerebral areas using the raw spatial transcriptomics information — a procedure termed parcellation — leads to indistinct maps.
That’s where the team’s AI-driven strategy demonstrated its value.
Extensive language models (LLMs), like ChatGPT, have enthralled and energized countless users with their aptitude to produce textual content from prompts. Fundamentally, these architectures operate by mathematically anticipating the interconnections between singular terms. Abbasi-Asl, in partnership with his doctoral candidate Alex Lee, conceived the AI system, designated CellTransformer, which instead scrutinizes how specific cells are situated adjacent to each other in the encephalon based on spatial transcriptomics insights.
The AI system transforms the spatial data, augmenting it with novel insights. “We’re creating a missing piece between spatial transcriptomics data and cerebral parcellation that connects the two,” Abbasi-Asl conveyed to Live Science. Per Abbasi-Asl, the novel dataset engendered by CellTransformer yields substantially more precise mappings that bear greater resemblance to recognized cerebral areas than manual delineation could, and it also pinpoints previously unlisted, more finely-grained regions.
Image 1 of 2
(Image credit: University of California, San Francisco)
This image showcases representations of the various cerebral sectors documented by CellTransformer.
(Image credit: University of California, San Francisco)
This 3D mapping illustrates a subset of the 1,300 areas within the murine encephalon that were engendered by CellTransformer.
The updated mapping encompasses approximately 1,300 parts of the mouse’s brain, resulting in a cumulative dataset of over 9 million cells. The team harmonized their findings with the Allen Institute’s Common Coordinate Framework (CCF), a high-definition mapping of the murine brain previously fashioned using manual annotation. Substantial uniformity existed between the AI-produced output and the gold-standard CCF, which fostered the team’s conviction that their conclusions possessed a high degree of precision.
CellTransformer effectively mapped established cerebral zones, such as the hippocampus, a pivotal center for memory. The instrument also charted cerebral regions that other mapping attempts had struggled to acquire information about, such as the midbrain reticular nucleus, (CK)which is situated in the uppermost segment of the brainstem and processes both sensory and motor data, while concurrently regulating sleep.
You may like
-
Blueprint of 600,000 neural cells revises the textbook on decision-making processes in the brain
-
Academics have just engineered a new AI simulating the human brain — it’s surpassing LLMs such as ChatGPT in logical thinking
-
‘Minibrains’ unearth mysteries of how vital cerebral cells emerge in utero
The data management underpinning CellTransformer is not solely tailored for cerebral matter, the authors emphasized.
“A comparable procedure could be employed with datasets presently surfacing from the heart, alternative body segments, and also from tissues amassed in disease prototypes as opposed to healthy prototypes,” Abbasi-Asl articulated.
RELATED STORIES
—In an inaugural effort, scientists integrate AI with a ‘minibrain’ to construct a hybrid computer
—AI ‘brain decoder’ possesses the ability to interpret an individual’s sentiments with merely a rapid brain assessment and virtually no instruction
—A novel 3D mapping charted with Google AI unveils a ‘mysterious yet gorgeous’ fragment of the human brain
The team also intends to assess CellTransformer utilizing human cerebral data — but while the murine brain comprises tens of millions of cells, our brains contain approximately 170 billion cells, encompassing 86 billion neurons. The sheer magnitude of the human encephalon, coupled with its more intricate composition, will render it more challenging to furnish an adequate volume of spatial data to nourish the AI.
If such data can be provided to CellTransformer, Abbasi-Asl posits the tools could process it, though. “We do believe that it could work on human data too,” he stated. “That’s another really important next step.”
RJ MackenzieLive Science Contributor
RJ Mackenzie stands as a science and health reporter with a nomination for an award. He holds academic degrees in neuroscience from the University of Edinburgh and the University of Cambridge. Transitioning to writing, he reasoned that he could best contribute to science from a writing desk rather than a lab setting. His reporting encompasses a wide array of subjects, spanning from brain-interface technology to shape-modifying materials science, and from the proliferation of deceptive conferencing to the relevance of newborn-screening initiatives. Previously, he served as a staff writer for Technology Networks.
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.
LogoutRead more
Blueprint of 600,000 neural cells revises the textbook on decision-making processes in the brain
Academics have just engineered a new AI simulating the human brain — it’s surpassing LLMs such as ChatGPT in logical thinking
‘Minibrains’ unearth mysteries of how vital cerebral cells emerge in utero
In ‘Secrets of the Brain,’ Jim Al-Khalili explores 600 million years of brain evolution to understand what makes us human
Science news this week: A key Atlantic current nears collapse, the world’s biggest iceberg shatters, and mouse brains rewrite neuroscience
China’s ‘Darwin Monkey’ is the world’s largest brain-inspired supercomputer
Latest in NeuroscienceSourse: www.livescience.com
