Summary: A multinational team used one of the world’s fastest supercomputers to build the most detailed digital simulation of a mouse cortex to date. The biophysically realistic model reproduces neuronal structure and activity, enabling virtual experiments on brain disorders, cognition, seizures and other neural processes.
Combining biological datasets from the Allen Institute with the extraordinary computing power of Fugaku, researchers can now observe neural activity across nearly ten million neurons and tens of billions of synapses. This milestone paves the way for larger whole-brain simulations and long-term efforts toward comprehensive digital brain models.
Key Facts:
- Massive scale: The simulation models ~10 million neurons, 26 billion synapses, and 86 interconnected cortical regions.
- Disease and mechanism research: The digital cortex enables virtual experiments on Alzheimer’s, epilepsy, brain waves, attention and cognition.
- Technical milestone: Demonstrates the feasibility of future full-brain, biophysically detailed simulations at scale.
Source: Allen Institute
Using the computational muscle of a top-ranked supercomputer, researchers have produced one of the largest and most detailed biophysically realistic animal brain simulations ever created.
This digital reconstruction of a whole mouse cortex gives scientists a new experimental platform: instead of relying solely on tissue samples and one-off experiments, they can simulate disease progression, test hypotheses about neural mechanisms, and evaluate interventions in a controlled virtual environment.
The cortex model reproduces both anatomical detail and electrical signaling, spanning nearly ten million neurons, 26 billion synaptic connections, and 86 functionally linked brain regions.
The project was powered by Supercomputer Fugaku, a Japanese flagship system capable of quadrillions of calculations per second, enabling simulation at an unprecedented level of biological fidelity and temporal resolution.
Researchers from the Allen Institute collaborated with Tadashi Yamazaki, Ph.D., of Japan’s University of Electro-Communications and three other Japanese institutions to build the model. A full account of the work will be presented at SC25, the leading supercomputing conference scheduled for mid-November.
With this model, scientists can investigate detailed questions that were previously impractical at scale: how pathological changes spread through networks in Alzheimer’s, how seizure activity propagates, or how specific patterns of brain waves influence attention and cognition. Virtual experiments can be repeated under controlled conditions to test mechanisms and interventions far more efficiently than with tissue alone.
“This shows the door is open. We can run these kinds of brain simulations effectively with enough computing power,” said Anton Arkhipov, Ph.D., an investigator at the Allen Institute involved in the project. “It’s a technical milestone that gives us confidence larger and more precise models are achievable.”
The collaboration pairs the Allen Institute’s extensive biological datasets with Fugaku’s computational capability. The Allen Cell Types Database and the Allen Connectivity Atlas supplied the cellular and circuit-level blueprints, while Fugaku enabled the computational realization of those data into a working cortex model.
How the whole-cortex simulation was built
Fugaku, developed by RIKEN and Fujitsu, can perform on the order of 400 quadrillion operations per second. Its architecture consists of many compute nodes arranged in units, shelves and racks—158,976 nodes in total—allowing enormous parallel workloads required by large-scale neuronal simulations.
Using the Allen Institute’s Brain Modeling ToolKit, the team converted experimental measurements into a digital blueprint. The neuron simulator Neulite translated mathematical descriptions of neurons into dynamic elements that spike, transmit synaptic signals and display realistic membrane dynamics.
The result behaves like biology in real time: the model captures neuron morphologies, branching dendrites, synaptic activation patterns, and the propagation of electrical signals across cell membranes. This level of biophysical detail allows researchers to probe mechanisms at cellular, synaptic and network scales.
“It’s a technical feat, but only the first major step,” said Yamazaki. “Precision in biophysically detailed models matters, and this work demonstrates we can achieve it at large scale.”
“Our long-term aim is to scale from single brain regions to whole-brain models, and eventually toward human-scale simulations as biological data and computational resources grow,” said Arkhipov. With advances in computing and richer experimental datasets, realistic full-brain models are moving from concept to feasible research tools.
This research was carried out by an international team including Laura Green, Ph.D.; Beatriz Herrera, Ph.D.; Kael Dai, B.Sc.; Rin Kuriyama, M.Sc.; and Kaaya Akira, Ph.D.
Key Questions Answered:
A: A biophysically realistic whole mouse cortex model with nearly 10 million neurons and 26 billion synapses.
A: It enables virtual experiments to model disease progression, test hypotheses about neural function, and evaluate interventions without relying exclusively on physical tissue experiments.
A: Fugaku’s massive computational power combined with high-resolution biological data from the Allen Institute’s neuroscience resources.
Editorial Notes:
– This article was prepared by a Neuroscience News editor.
– The journal paper was reviewed in full.
– Additional context was provided by editorial staff.
About this neurotech and brain mapping research news
Author: Liz Dueweke
Source: Allen Institute
Contact: Liz Dueweke – Allen Institute
Image: The image is credited to Neuroscience News
Original Research: The findings will be presented at SC25