(Content from the 2025 academic year onwards)
We are a laboratory for neural circuit simulation.
Our lab is dedicated to the science and technology driven by information, focusing on brain neural circuits as our research target, with numerical computation and High-Performance Computing (HPC) as our core technologies. The research goal is to construct a biologically detailed model of the brain’s neural circuits, which we refer to as a digital copy of the brain. To achieve this, we are developing a simulator called Neulite for simulating these biologically detailed models.
In a biologically detailed model, individual neurons possess characteristic spatial shapes; current diffuses within the cell, leading to changes in calcium concentration, which in turn causes numerous ion channels to open and close, resulting in highly nonlinear spike responses. The reasons why we require this level of model resolution are:
- Raw experimental data (anatomical and electrophysiological) can be directly incorporated.
- We aim to unravel the causal relationship between microscopic changes in neural circuit properties and various macroscopic phenomena in the brain (e.g., abnormal brain waves, behavioral changes).
- Utilizing the spatial shape dramatically enhances the information processing capability of a single neuron.
- It truly necessitates the performance of supercomputers.
As the saying goes, “God is in the details,” and we believe this holds true for biologically detailed models as well.
Currently, we have succeeded in running a biologically detailed model of the entire mouse cerebral cortex on the full scale of the Fugaku supercomputer. Since supercomputer performance increases by a factor of 100 every 10 years, we anticipate that within the next 20 years, we will be able to construct a micro-scale resolution digital copy of the entire human cerebral cortex, or even the whole human brain including the cerebellum.
We believe that using biologically detailed models can contribute to:
- Elucidating the information processing mechanisms of neural circuits.
- Reproducing macro-scale neurological/psychiatric disorders stemming from microscopic abnormalities in neural circuits.
- Developing low-power AI technology inspired by the brain.
Our Research Focus and Specialization
Our current core interest is the development of the simulator. It is not research in so-called computational neuroscience that utilizes the simulator. There are several laboratories in Japan and a considerable number internationally that study computational neuroscience. While we possess international competitiveness, particularly concerning the cerebellum, this is not necessarily unique research.
In contrast, in the field of simulator development, as of 2025, we are the world champions, and this research can only be done by us. Therefore, to maximize the laboratory’s advantage, we are deliberately specializing in simulator development.
This is an important point for students from other universities who wish to join our lab. The research areas available here include:
- Introduction of low-precision/mixed-precision computation schemes.
- Development of parallel computation techniques for reaction-diffusion systems.
- Development of training methods for large-scale networks.
We especially welcome students from the HPC domain who wish to venture into brain research.
Past and Current Student Research
While some enrolled students are conducting traditional computational neuroscience research, we will refrain from starting new projects in that area going forward. If one simply wants to work with spiking neurons, all the necessary know-how is compiled in the “black book” (referring to a comprehensive guide/textbook in Japanese); reading it should enable such work in any lab. We are moving beyond that.