Group Leader
Kenji Doya
Professor, Neural Computation Unit, Okinawa Institute of Science and Technology Graduate University
The digital brain is a mathematical model of the brain reconstructed by integrating anatomical and physiological data of the brain. The software program for creating digital brains will be made publicly available and the "Brain Data Integration Platform," which will allow connections to brain databases, model building, and simulation in a cloud platform. In collaboration with other groups of the Core Organization and other institutions, digital brains will be created and used in various simulation experiments. The digital brains will be used to test the models of brain functions, such as reinforcement learning and Bayesian inference, and to explore interventions for mental and neurological disorders.
Subprojects
4A
Interspecies Conversion and Integration of Brain Function Maps
Shin Ishii / Henrik Skibbe
Professor, Department of Informatics, Kyoto University Graduate School of Informatics / Unit Leader, Brain Image Analysis Unit, RIKEN CBS
The "Interspecies Brain Integrative Platform," a tool to compare the functional circuits between multiple species (humans, macaques, marmosets, and possibly mice), using diffusion and structural MRI data will be developed. Using this platform and the brain activity data that will be generated by the Core Organization, the similarities and differences of neural circuits that produce spontaneous and functional activities will be examined across species.
4B
Integration and Expansion of Human Brain Databases
Saori C. Tanaka
Department Head, Department of Neural Computation for Decision-making, Brain Information Communication Research Laboratory Group, Advanced Telecommunications Research Institute International
In an effort to build a data infrastructure that supports the development of the digital brain, human brain MRI databases of neuropsychiatric disorders will be expanded and consolidated. Specifically, technical support will be provided to link new datasets generated in this program to existing brain databases (e.g., Brain/MINDS Beyond, Integrative Brain Research, and IBISS) at the metadata level. The combined datasets will be used to develop a framework for cross-species data querying.
4C
Multilevel Data-Driven Model Building
Kenji Doya
Professor, Neural Computation Unit, Okinawa Institute of Science and Technology Graduate University
New mathematical methodologies and software programs will be developed for creating digital brains at multiple levels ranging from local circuits to the whole brain by integrating a variety of data about the structure and activity of the brain as well as perception and actions. The developed software tools will be made publicly available and a cloud-based system will be launched to enable brain modeling using databases and running simulations ("Brain Data Integration Platform").
4D
Reproducing Brain Functions and Pathology Through Simulation and Data Assimilation
Jun Igarashi
Senior Scientist, High Performance Artificial Intelligence Systems Research Team, RIKEN Center for Computational Science
A neural circuit simulator that can incorporate connectome, transcriptome, electrophysiology, and other brain data will be developed and neural circuit models will be created from rodent and primate brain data. Through neural circuit simulation using data assimilation and parallel processing techniques, it will be tested whether neural activities in healthy and diseased states can be reproduced. The results will clarify the brain's information processing mechanisms and the pathogenetic processes of brain disorders.
4E
Development of Methods for Analyzing Brain Computational Mechanisms
Takuya Isomura
Unit Leader, Brain Intelligence Theory Unit, RIKEN CBS
Using the digital brain, the brain's computational mechanisms of sensory perception, motor control, and learning will be explored. Based on the concept of Bayesian mechanics that considers general dynamical systems as Bayesian inference, new analysis methods will be developed. Using the methods, it will be tested whether the circuit architectures and dynamics for predictive coding, active inference, or control as inference, which constitutes the Bayesian brain, emerge in the digital brains that are constructed from the data in a bottom-up fashion.
4F
Neuropsychiatric Disease Modeling
Saori C. Tanaka
Department Head, Department of Neural Computation for Decision-making, Brain Information Communication Research Laboratory Group, Advanced Telecommunications Research Institute International
The mechanisms underlying neuropsychiatric diseases will be investigated by using computational models and linking patients' multilevel data that will be collected in this program (e.g., invasive recordings, stimulus responses, macroscale brain activities, behavioral data, and treatment records).