قالب وردپرس درنا توس
Home / Innovative / Why do you need a supercomputer to map a mouse brain?

Why do you need a supercomputer to map a mouse brain?



In a 25,000-square-foot space at Argonne National Laboratory, Theta, one of the world's most impressive supercomputers, applies its incredible computing power to the largest amount of data ever recorded or analyzed. Information that researchers hope will someday contribute to our understanding of intelligence.

And in this case, all of this data fits in the skull of a mouse.

Theta is currently mapping the structures of mouse brains using a dataset collected in small numbers from Narayanan "Bobby" Kasthuri, a neuroscientist at the Argonne National Laboratory and assistant professor of neurobiology at the University of Chicago. When the entire set, soup to nuts, is procured, the final result is expected to be a million terabytes, a tremendous amount that is impossible to grasp raw information.

Kasthuri explains his ambitions for this massive amount of data after storming him on Drexel Avenue to escape the cold on the University of Chicago's medical campus near his lab. For him, the future hopefully holds an unprecedented understanding of the brain.

In Khethta's lab, colored diagrams are created with color blobs gliding past each other with the cyclic psychedelia of a Broad City title. What they depict is as trippy as it looks; small slices of the mouse brain, pinned and marked in color, to represent the brain structure on a tiny level. The splashes of color are neurons and synapses, the paths through which the mind moves, and each of them is documented in a nanometer-sized map called a connectom.

Researchers believe a complete connectome could help solve the secrets of cognition and mental health, and deepen our understanding of the differences and similarities between the brains of different organisms. At the basic level, knowledge about the working units of the brain is conveyed, which is crucially lacking in current research.

Currently, Kasthuri is interested in making and comparing sub-connec- tions from different types of animal brains. At the moment, the lab is working on sections related to addiction in mice. Compared to a non-addictive mouse connectome, a mouse-dependent-to-cocaine connectom could find out which neurons are affected by the addiction.

"We have already established – it appears that there are structural changes in the search brain," Kasthuri says.

Brains on a Conveyor Belt

Studying the structure of a mouse brain is a delicate process with great potential for error. The brains are picked as quickly as possible by the mouse, which has been preserved with aldehyde fixatives in the race against the damage of death. For the scanning electron microscope (SEM) in Argonne, a part of the brain is immersed in heavy metal stains. After the sample has been dehydrated and plasticized, it is cut with a dia cutter similar to a diamond knife with a width of several atoms. In typical cutting systems, cutting and moving the samples can lead to imperfections that can be magnified when the supercomputer is involved.

These tiny flaws can be a major problem when researchers analyze the data, says Animashree Anandkumar, Bren Professor of Computer and Mathematics at Caltech is not involved in the project. "These distortions can lead to incorrect correlations," she says.

Kasthuri's solution is simple – a proprietary delivery system that quickly slips brain slices and minimizes human error. After traveling on the conveyor belt, samples are scanned by the SEM, resulting in image data stacks. The individual neurons, synapses, and other structures are recognized by the shape, tracked, and then stained as in Photoshop, a boring but crucial work done by students such as Anastasia Sorokina and Katrina Norwood, graduates of Kasthuri's lab.


Kasthuri Laboratory Members Anastassia Sorkina, Katrina Norwood, and Rafael Vescovi
B David Zarley

Unfortunately, it takes an infinite amount of time for each human to digitally color each structure, and analysis of the results leads to bottlenecks for connectomics. Kasthuri's dream – a complete mouse brain; the "mouseshot" – would be virtually impossible with people alone. He has the bill. Any person on Earth who works perfectly, eight hours a day, six days a week, would take another 500 to 1,000 years to complete the project. One hundred, if you could design anyone who ever lived.

A coloring book for computers

The answer is an algorithm. In particular, flood-filling networks developed by Google AI and the Max Planck Institute for Neurobiology in Germany.

"If these datasets are billions and billions and hundreds of billions of pixels, this analysis by humans is impossible," says Virus Jain of Google AI.

According to Jain, one of the architects of the algorithm used, the data flooding algorithm approximates how a person would make a coloring book. It begins with a particular structure – such as a particular neuron – that fills before it passes to others. Theta comes into play here.

Haritha Siddabathuni Som, Team Leader of the Argonne National Laboratories Leading Computing Facility, lists the impressive values ​​of the supercomputer. It's the most powerful supercomputer in the system, occupying 24 server racks, and if it does not consume any mice minds, it works on other huge datasets, including some of CERN's Large Hadron Collider, who are delving into the secrets of particle physics.


Theta Supercomputer of the Argonne National Laboratory
B David Zarley

Theta's Intel Cray XC40 hardware is capable of delivering 11.69 petaflops, which is really crazy for laymen . Humans can analyze a cubic micron of mouse brain in about 2 minutes. In a mouse brain there are a trillion cubic meters. While mankind would take centuries to complete the task, Kasthuri believes that theta can use the algorithm and do it in just five years.

The flood flooding algorithm uses Theta's horsepower to locate, stain, and assemble the mouse brain data and sends it to Argonne's analysis and visualization cluster, Cooley, which produces the acid triponnones. The datasets are open source and are available for everyone to view and study.

The algorithm's approach is an order of magnitude better than previous options. But even with a state-of-the-art algorithm and a powerful supercomputer, the progress toward a whole mouse connectome is still slow.

A complete connectome is rare; In the late eighties, for example, there was a worm C. elegans. The next whole brain will probably be the fruit fly, a scientific battle. The FlyEM project at Janelia Research Campus aims to publish a complete connectome of about one-third of the fruit fly brain within a year, says Stephen Plaza, a project scientist, over the phone.

About 30 years separate the whole union of the C. elegans worm from the fruit fly brain; By comparison, Kasthuri expects his mouse brain to be completed within five years of the fly. The financing is still pending. The ultimate goal is bold, even when compared to the mouseshot: a complete connection of the human brain. This effort will take more time, and an even larger and more powerful supercomputer, Aurora 21, which is currently being built in Argonne.

Kasthuri leans back when he talks about the possibilities that a full human connectome offers. He imagines the ability to find and correct problems caused by mental disorders and traumatic brain injuries. He imagines proving that knowledge is the result of building connections, or how a marble sculptor degrades it (he likes the latter). He and the other researchers in this field are coming up with answers that will naturally arouse more curiosity.

"We always find that looking at the Connectome basically seems to be a Pandora's box," says Plaza of the FlyEM team. "Not for answers, but for more questions."


Source link