The supercomputer has made it possible to create one of the most detailed brain simulations from open data. Model of the cerebral cortex (outer layer of the brain) of a mouse It was implemented by the Allen Institute in Seattle, USA. It contains approximately 10 million neurons, 26 billion synapses, and 86 interconnected areas in the brain.
The results were announced on the 17th. This high level of detail was made possible by the Japanese supercomputer Fugaku, developed by the Riken Center for Computational Science and technology company Fujistsu.
The machine can perform 400 quadrillion operations per second. For perspective, calculating the number per second to reach 400 quadrillion would take 12.7 billion years. This ability is made possible by approximately 160,000 nodes, which are small parts of the computer that are grouped into layers and allow huge amounts of data to be managed.
Diego Szczubak, a professor in the Department of Neuroscience at the University of Pittsburgh in the United States, believes that the work done by the Allen Institute is impressive and that brain simulations usually involve thousands or tens of thousands of synapses, rather than billions.
The simulation can capture the structure and behavior of brain cells, the branches that give off neurons, the activation of synapses, and electrical signals. In the long term, the goal is to build complete models, even for humans.
The idea is that researchers can use the model to understand diseases, such as epilepsy and Parkinson’s, and how these problems can arise even before symptoms appear.
Simulations at this level of detail require a large volume of data. In the case of Fugaku’s simulation, the researchers used databases from the Allen Institute itself, a non-profit entity that aims to answer key questions in biological sciences. One of them retains information about the morphology and electrical properties of neurons.
The brain remains a mystery today. To understand the function of each part, it is necessary to study how regions communicate with other regions. Using simulations to understand the brain has the advantage of being able to study the interaction between brain regions, preserving details of neuronal activities, Szczupak explains.
When analyzing an animal’s brain while performing tasks, it is not possible to monitor many areas at the same time. “The best we can achieve (through animal analysis) is still very limited at present compared to simulating all the neurons in the brain,” says the professor.
The simulation created by Fugaku allows synapses occurring at the same moment to be observed in great detail, something that is not possible in a living animal. Thus, scientists can study how the signal that originates in a certain part of the brain spreads to the remaining part.
A video of the simulation released by the Allen Institute shows a mouse cortex at subcellular level resolution. Neurons vary in color depending on the area of the cortex, and are characterized by light when they are active.
But Szczupak highlights that it is not possible to reproduce reality in models. There are also other limitations in the simulations, such as the lack of variation in the anatomy and flexibility of the organs. The last concept relates to the changes that neurons undergo even when they have already formed, as well as other elements that make up the brain.