Mouse brain research is helping us better understand human minds

Jul 08, 2024

[[read-time]] min read

Ari Marini

Keyword contributor

Google researchers precocious unveiled nan largest, astir elaborate representation of nan quality encephalon yet. It described conscionable 1 cubic millimeter of encephalon insubstantial — nan size of half a atom of atom — but astatine precocious capable solution to show individual neurons and their connections to each other, and required 1.4 petabytes of information to encode.

Although it’s only a mini sliver of nan brain, nan representation led to respective astonishing discoveries. “For example, we recovered immoderate of nan wires will wrap themselves into these elephantine knots,” Google Research Scientist Viren Jain says of nan neurons. “We person nary thought why — nobody's ever seen it before.”

Now, Viren and his squad person sewage mice connected nan brain. And for bully logic — these mammals whitethorn thief lick mysteries astir our minds that person eluded america since our beginnings. Mysteries like: How are memories stored and retrieved? How do we admit objects and faces? Why do we request truthful overmuch sleep? And what goes incorrect successful Alzheimer’s and different encephalon diseases?

“One logic we don’t person answers to these questions is that we don’t yet person nan information we request successful bid to study nan brain,” Viren says.

The quality encephalon has astir 86 cardinal neurons connected to each different by much than 100 trillion synapses that alteration you to think, feel, move and interact pinch nan world. By creating a representation of these neural connections — known arsenic a “connectome” — we tin unlock caller knowing astir really our brains work, and why sometimes they don’t.

To build elaborate maps astatine nan synaptic level, researchers request to image nan encephalon astatine nanometer solution and activity pinch monolithic amounts of data. It’s a important method situation that requires continued invention successful imaging techniques, AI algorithms and information guidance tools. That’s why, 10 years ago, Google Research formed its Connectomics team.

Over nan past decade, nan squad has developed technologies to more efficiently process, analyse and stock data, enabling researchers to dramatically standard up advancement connected knowing nan brain. For example, they introduced flood-filling networks, which replaced nan manual effort of coloring successful cells crossed encephalon images by utilizing instrumentality learning to automatically trace nan paths of neurons done layers of tissue. Building connected this, their SegCLR algorithm automatically identifies chopped parts of cells and compartment types wrong these networks. And they created package specified arsenic TensorStore and Neuroglancer which helps store, process and visualize ample multidimensional images and volumes.

Still, mapping nan full quality encephalon connectome would require gathering and analyzing arsenic overmuch arsenic a zettabyte of information (one cardinal terabytes), which is beyond nan existent capabilities of existing technologies. “If we were to representation nan full quality encephalon correct now, it mightiness return billions of dollars and hundreds of years,” Viren says.

Instead, researchers attraction connected either mapping ample portions of nan brains of mini animals, aliases mini pieces of encephalon insubstantial from ample animals. In 2020, nan Connectomics squad mapped half of a consequence fly’s brain, revealing nan connections among 25,000 neurons. Through collaborations pinch researchers successful nan field, they’ve besides created connectomes for portions of nan brains of nan zebra finch and zebrafish larvae. And successful May, nan aforementioned representation of 1 cubic millimeter of quality encephalon insubstantial was published successful Science.

Thousands of researchers astir nan world person utilized nan datasets from these projects, resulting successful hundreds of published discoveries.

Now, nan Connectomics squad is moving pinch partners astatine Harvard, Princeton and elsewhere to representation nan mouse’s hippocampus — nan portion of nan encephalon responsible for encoding memories, attraction and spatial navigation, and representing 2-3% of nan full rodent brain.

Without nan clip aliases exertion to representation nan full quality brain, analyzing a rodent connectome is nan adjacent champion thing. It’s mini capable to beryllium technically feasible and could perchance present insights applicable to our ain minds. “When you look astatine a rodent encephalon successful nan particle microscope, it looks precisely for illustration a quality brain. It is, successful fact, a miniature type of a quality brain,” Jeff W. Lichtman, a professor of molecular and cellular biology astatine Harvard, says. That's why scientists often usage mice to study quality encephalon disorders.

Mice are conscionable nan latest connectome frontier; neuroscientists person been moving for decades to representation progressively larger and much analyzable brains. The first connectome was for a worm’s encephalon — published successful 1986, it took 16 years to map.

Even though a rodent encephalon is 1,000 times smaller than a quality brain, mapping 1 is still an immense method challenge. The dataset from 1 rodent encephalon connectome astatine nanometer solution could beryllium nan largest biologic dataset ever collected, estimated astatine astir 20,000-30,000 terabytes.

“So not conscionable acquiring nan data, but moreover conscionable storing and accurately processing each that information is simply a awesome challenge,” Viren says. “But that’s been our unsocial publication to nan field: processing devices that push nan authorities of nan creation successful accuracy and past really applying them astatine standard to larger and larger datasets.”

If successful, nan Connectomics team's rodent encephalon task will beryllium nan first clip scientists person mapped portion of a mammalian hippocampus. It’s besides nan largest-ever chunk of immoderate encephalon researchers person attempted to map.

“Fundamental investigation generates bonzer value,” Viren says. “What gets maine excited is that 1 day, we’ll person a precise knowing of really we shape memories and what underlies intelligence disorders aliases diseases. But successful bid to do that, we person to create this loop of exertion that would person been unimaginable arsenic precocious arsenic 2 decades ago.”