A Single Cubic Millimeter of Brain Tissue May Have Just Changed Neuroscience Forever

“Hearst Magazines and Yahoo may earn commission or revenue on some items through these links.”

A new package of papers examines the largest map yet of mammalian brain tissue.
The map shows one cubic millimeter worth of neurons in the visual cortex of a mouse.
Many brain functions, particularly the senses, are similar across different mammal species.

Scientists have mapped an unprecedentedly large portion of the brain of a mouse. The cubic millimeter worth of brain tissue represents the largest piece of a brain we’ve ever understood to this degree, and the researchers behind this project say that the mouse brain is similar enough to the human brain that they can even extrapolate things about us. A cubic millimeter sounds tiny—to us, it is tiny—but a map of 200,000 brain cells represents just over a quarter of a percent of the mouse brain. In brain science terms, that’s extraordinarily high. A proportionate sample of the human brain would be 240 million cells.

Within the sciences, coding and computer science can sometimes overshadow the physical and life sciences. Rhetoric about artificial intelligence has raced ahead with terms like “human intelligence,” but the human brain is not well enough understood to truly give credence to that idea. Scientists have worked for decades to analyze the brain, and they’re making great progress despite the outsized rhetoric working against them.

That said, artificial intelligence designed for specific tasks is essential to research like this. In a series of eight papers in the peer reviewed journal Nature, the team behind the Machine Intelligence from Cortical Networks (MICrONS) project—hailing from the Allen Institute, Baylor College of Medicine, and Princeton University—described how they used machine learning to “reverse engineer the algorithms of the brain.”

The field in which scientists map the brain and other parts of the nervous system (of humans or any other creature) is called connectomics. The term comes from the same suffix as in biome or genome, referring to a complete picture or map of something. This work expands on the connectome—which is only the physical map—by adding data about each neuron’s function.

In one of the team’s papers, the researchers were able to make an overall classifying system to cover 30,000 neurons by their different shapes, or morphologies. These neurons are excitatory, meaning they’re involved with transmitting messages in the brain. The alternative to excitatory is inhibitory, which is circuitry that stops a message from being passed, like an insulator.

Inhibitory neuron shapes are better understood, partly because their shapes can be separated into diverse (but discrete) groups. In this study, scientists used machine learning to help classify excitatory neurons, which seem to need a more complicated classifying system. By turning the neurons into measurements, observations,and layers, the scientists could then use statistical methods to find how often certain types or qualities of these cells appeared. This may sound like an oxymoron, but code can generalize more precisely than human scientists are able to.

“(1) Superficial L2/3 neurons are wider than deep ones; (2) L4 neurons in V1 are less tufted than those in HVAs; (3) the basal dendrites of a subset of atufted L4 neurons in V1 avoid reaching into L5; (4) excitatory cortical neurons form mostly a continuum with respect to dendritic morphology, with some notable exceptions.”

The conclusion about a continuum is really important. Having categories for neurons can be and has been useful in studying the brain, but computing power can deepen this understanding and add a great deal of nuance. With more information, we can turn broad types into something more individualized.

Another paper in the set found confirmation of an existing theory that “like connects like” within neuron structures. Neurons that perform certain tasks in the visual cortex of the mouse brain reach out and link up with each other, whether they’re adjacent or layers apart. Because of the size of this dataset, the scientists were able to extend this established theory into further-away parts of the brain region. And since even this large mapping of brain tissue is still very incomplete, the number of “like” neurons is likely even higher in reality.

The data and maps from this project are available for the public to check out by following the instructions on their website. It’s wild that you don’t even have to download anything—you can map the brain using your web browser.

You Might Also Like