Extensive new data sets extend the boundaries of neuroscience.

Approximately 300,000 mouse neurons were recorded during this Allen Institute release. The difficulty now lies in deducing what actionable insights may be drawn from this mountain of information.

IN MOST INTRODUCTORY NEUROLOGY CLASSES, STUDENTS WATCH THE SAME VIDEO. It doesn't look like much, just a bar of light revolving and moving over an otherwise dark screen, with some distant fireworks noises in the background. Uninteresting until you find out that each pop is the activity of a single neuron in a cat's brain as it watches the bar on the screen move. The popping culminates in a flurry of activity when the bar reaches a certain point and orientation. This neuron has made its feelings on the bar abundantly apparent.

In the 1960s, David Hubel and Torsten Wiesel conducted the experiment depicted in the movie, allowing researchers to draw meaningful conclusions regarding the visual system. Neuroscientists have been inserting thin metal electrodes into mice, finches, and monkeys' brains for decades to observe individual neurons and discover their triggers. Some neurons are activated in response to particular colours or forms or the position of an object in space, the orientation of a person's head, the entirety of a face, or a specific feature within a face.

Anne Churchland, a professor of neuroscience at the University of California, Los Angeles, remarked that "everyone constantly wanted more neurons," even though single-cell analysis has demonstrated the brain to be a potent engine. More data can be gathered from an experiment is a contributing factor. However, researchers also hit analytical roadblocks while attempting to study individual neurons. Prefrontal cortex neurons respond to a wide array of stimuli (visual characteristics, tasks, decisions) that scientists can't identify their function, at least on an individual level. However, Hubel and Wiesel recorded activity in the primary visual cortex, a region located at the very rear of the brain; only a small percentage of neurons fire when the animal views oriented bars.

The methods developed by Hubel and Wiesel did not simultaneously allow for examining more than a few neurons. However, engineers have continued to push that limit, and in 2017 they created Neuropixels probes. A single probe, measuring only one centimetre in length and composed of silicon, can listen to hundreds of neurons simultaneously and is small enough that neuroscientists can insert multiple probes into an animal's brain. To record simultaneously from eight separate parts of the mouse visual system, researchers at the Allen Center, a nonprofit research institute founded by Microsoft co-founder Paul Allen, used six Neuropixels probes. The institute published its findings in August, covering the behaviour of over 300,000 neurons in 81 mice. Those who wish to use the data are welcome to do so at no cost.

The publication of the data collection, which is three times the size of the previous record holder, will allow scientists to study the coordinated behaviour of massive networks of neurons for the first time. Because of its unparalleled scale, this study may provide insights into hitherto inaccessible facets of the human mind. Shawn Olsen, a key investigator on the Allen Institute project, said, "We want to understand how we think and see and make decisions." "This just does not occur at the level of individual neurons."

Exactly how to process all that information is the current obstacle. Massive data sets challenge even the simplest tasks, like sharing or downloading. However challenging the analysis may be, many researchers find it well worth working with large data sets to learn about the brain in its natural environment.

When Hubel and Wiesel looked at the brain, they saw a factory with rows and columns of specialized neurons working together to complete a task. A red balloon will elicit different responses from neurons that detect red and those that detect circles when you show it to someone. However, the brain is so intricately connected that no one neuron acts alone. Therefore this method never worked. Scientist and Columbia University professor Stefano Fusi claims the brain does not examine individual neurons. In a single brain cell, thousands of neurons might gaze at each other. And thus, let's share the same point of view."

Some brain parts, like the prefrontal cortex, function like a workshop with experienced artisans. When potters with different skills unite, they can produce complicated and beautiful objects. We benefit from this variety, and it's probably crucial to the sophisticated reasoning and problem-solving abilities that set us apart. (Fusi showed that when neuronal populations display a rich diversity of responses to varied conditions, monkeys tend to perform better on a memory task; this was verified in research of the prefrontal cortex.) Conversely, highly specialized neuron populations are rigid and limited in what they can accomplish, much like a factory's assembly line.

However, assembly lines can be understood by everyone. Individual process steps can be analyzed for their specific contributions to the final result. Prefrontal cortex neurons can't be understood without the rest of the brain. Humans need sophisticated mathematical skills to make sense of these group activities. It's not something you can picture, Fusi explains.

To provide this kind of visualization, neuroscientists employ a method called "dimensionality reduction," in which they take data from hundreds of neurons and, using ingenious linear algebraic techniques, characterize their activity using just a few variables. In the 1990s, psychologists categorized people into five distinct personality traits: openness, agreeableness, conscientiousness, extroversion, and neuroticism. They found that knowing how someone fared on those five characteristics could effectively predict how that person would fare on hundreds of questions on their personality.

However, the factors derived from brain data cannot be reduced to a single adjective, such as "openness." Similar to motifs, these patterns of brain activity can be found across entire networks of neurons. The axes of a graph can be defined by a small number of these motifs, with each point representing a distinct set of circumstances based on the combination of these motifs.

There are drawbacks to simplifying information from thousands of neurons into a few variables. Like how certain buildings disappear from a 2D photograph of a 3D cityscape, so too can compressing a complex set of neuronal data into only a few dimensions lose a tremendous deal of complexity. To examine hundreds of individual neurons would be an overwhelming task, but working in a few dimensions is far more doable. Researchers can observe the neurons' changing behaviour over time by plotting their activity evolution against axes defined by the motifs. The motor cortex, where researchers had long been baffled by the perplexing and unexpected responses of single neurons, has significantly benefited from this method. However, when the neurons are observed as a group, they follow predictable patterns that are frequently cyclical. Different facets of motion are reflected in different characteristics of these trajectories; for instance, the location of these trajectories is correlated with their velocities.

Olsen claims that scientists will utilize dimensionality reduction to decipher meaningful patterns in the data. We can't examine each neuron individually," he explains. We need statistical and machine learning techniques to help us make sense of massive datasets.

It's true that this line of inquiry is in its infancy and that experts have a hard time agreeing on what the patterns and trajectories indicate. John Krakauer, a professor of neurology and neuroscience at Johns Hopkins University, adds, "People dispute all the time about whether these phenomena are factual." Can we believe them?" How quickly do they translate [into single-neuron responses]? Not as solid and real, if that makes sense.

The availability of large-scale data sets like the Allen Institute's will undoubtedly make it easier to bring these trajectories down to earth, as stated by Churchland. The institute's resources and sizeable research team are ideal for generating the volumes of data needed to put these instruments through their paces. Olsen compares the institute to an astronomical observatory, noting that while no single lab could afford its cutting-edge technology, the whole scientific community benefits from and contributes to the center's extensive experimental facilities.

He adds the Allen Institute is currently piloting a system where scientists from the scientific community can advise what stimuli animals should be presented with and what tasks they should be completing. At the same time, thousands of their neurons are recorded. As the ability to record brain activity improves, scientists are developing increasingly complex and lifelike experimental paradigms to study how neurons react to complex challenges that truly test their skills. Fusi argues that presenting the cortex-oriented bars is insufficient if we want to learn about the brain. We have to go through this.