Detailed virtual brain simulations is changing how we study the brain (alleninstitute.org)

Researchers led by the Allen Institute and Tadashi Yamazaki used Japan’s Fugaku supercomputer to build one of the largest biophysically realistic brain simulations to date: a whole mouse cortex model with almost 10 million neurons, ~26 billion synapses and 86 interconnected brain regions. Using the Allen Institute’s Brain Modeling ToolKit, cell- and connectivity data from the Allen Cell Types Database and Connectivity Atlas were converted into functioning circuits; the Neulite simulator turned biophysical equations into spiking neurons that reproduce dendritic morphology, synaptic activations and membrane voltage dynamics. The run leveraged Fugaku’s ~158,976-node architecture and more than 400 quadrillion operations per second to simulate detailed, real-time neural activity. Results will be presented at SC25. This milestone matters because it shifts many neuroscience experiments from slow, singular wet-lab trials to fast, repeatable in-silico experiments: researchers can now probe disease spread (Alzheimer’s, epilepsy), seizure propagation, brain rhythms and circuit-level hypotheses, and safely test interventions in a controlled digital tissue. Technically, it demonstrates that full-cortex, biophysically detailed simulations are tractable at scale and paves the way toward larger whole-brain—and eventually human—models. The work underscores both the power of combining high-resolution biological datasets with exascale-class compute and the new opportunities and challenges for validation, interpretation and translational use.