Past Seminars

Abstract: Healthy Brains, Healthy Lives (HBHL) at McGill University is a $100M initiative to apply big data approaches in brain research. HBHL supports >250 independent researchers in studies of neurodegeneration, developmental/psychiatric disorders and brain plasticity. To support this community's work, the Evans laboratory, (MCIN) has developed a data ecosystem that incorporates multi-modal databasing (LORIS), high-performance computing (CBRAIN) and open data-sharing (CONP) platforms. This infrastructure supports many national (e.g. EEGNet, CCNA, C-BIG) and international (e.g. GBC, IBIS, HBCD, HIBALL) initiatives. The talk will provide an overview of this data ecosystem and its applications.

Abstract: Understanding the brain requires linking microscopic biophysical properties (e.g., ion channel dynamics and dendritic morphology) to emergent macroscopic phenomena (e.g., neural oscillations and network dynamics). While biologically detailed models are instrumental for this mechanistic insight, their inherent computational cost is substantial. While conventional simulators like NEURON and Arbor are highly effective for general use, in simulating models at the extreme scale, they could become inefficient. This is primarily in terms of their scalability and optimization to fully leverage modern High-Performance Computing (HPC) architectures.

We introduce Neulite (https://numericalbrain.org/neulite/), a light-weight biophysically detailed neural circuit simulator. It is architecturally defined by two core components: a frontend compliant with the Allen Institute’s Brain Modeling ToolKit (BMTK), and a portable numerical kernel. The frontend facilitates biological plausibility and reproducibility through the utilization of standardized data. The kernel, which can be specifically tuned for different computing architectures, allows us to overcome the limitations of conventional simulators through optimized, domain-specific algorithms.

Neulite has been utilized to successfully execute the Allen Institute’s whole mouse cortex model on the Supercomputer Fugaku. We used the entire Fugaku system to simulate 9 million biologically detailed neurons and 26 billion synapses, demonstrating a significant scale of computation. This work is the result of an international collaboration with the team of Dr. Anton Arkhipov at the Allen Institute, where their comprehensive model met our high-performance simulation technology, with support from key domestic contributors (RIKEN R-CCS, RIST, and Yamaguchi University). Neulite is therefore a valuable tool for achieving data-driven, large-scale modeling and advancing the understanding of how cellular properties influence overall neural circuit function.

The content of this talk has been published in a conference paper [1].

[1] Rin Kuriyama, Kaaya Akira, Laura Green, Beatriz Herrera, Kael Dai, Mari Iura, Gilles Gouaillardet, Asako Terasawa, Taira Kobayashi, Jun Igarashi, Anton Arkhipov, Tadashi Yamazaki (*: equally contributed). Microscopic-Level Mouse Whole Cortex Simulation Composed of 9 Million Biophysical Neurons and 26 Billion Synapses on the Supercomputer Fugaku. in The International Conference for High Performance Computing, Networking, Storage and Analysis (SC ’25), November 16–21, 2025, St Louis, MO, USA. ACM, New York, NY, USA, 11 pages. doi: 10.1145/3712285.3759819.

Abstract: Understanding how brain-wide neural circuits are genetically organized remains one of the fundamental challenges in neuroscience. Roger Sperry's classical chemoaffinity theory proposed that molecular gradients provide positional cues for axonal wiring, yet its application has been largely limited to localized sensory systems. Here, we introduce SPERRFY (Spatial Positional Encoding for Reconstructing Rules of axonal Fiber connectivitY), a data-driven framework to examine whether Sperry's concept can be generalized to the entire brain. By integrating mesoscale connectomic data with spatial transcriptomic maps from the Allen Mouse Brain Atlas, SPERRFY applies canonical correlation analysis (CCA) to identify latent correlated structures between gene-expression and connectivity spaces, thereby inferring positional gradients that may reflect molecular constraints underlying long-range axonal organization. This framework bridges molecular and anatomical levels of organization, providing a quantitative basis for reinterpreting Sperry's chemoaffinity theory in the context of whole-brain connectomics.

Abstract: When we collaborate with others to tackle new challenges, we anticipate the roles that each team member will play, consider our own role based on these predictions, and adjust our behaviour accordingly. Because each member differs in experience and skills, it is necessary to flexibly adapt the strategy used for prediction. For example, when interacting with beginners who have less career experience, we can make reasonable predictions about their performance by projecting adjusting introspection ('social metacognition'). In contrast, the same strategy cannot be applied to experts with more extensive experience. In this talk, I will introduce our latest research addressing this issue. Specifically, we found that the anterior lateral prefrontal cortex (area 47) is engaged when predicting the thoughts of beginners through social metacognition, whereas the temporoparietal junction (TPJ) is recruited when predicting the thoughts of experts based on heuristics. The lecture will begin with an overview of the foundations of metacognition and then present the studies that led to these findings.

Abstract: Perceptual decisions require that neural populations encode information about the sensory environment and that other downstream populations read it out to inform behavioral outputs. Here we present our computational work to provide methods that individuate and tease apart the contribution of these two neural operations to the formation of perceptual decisions. We exemplify these methods with the study of several datasets from sensory and parietal cortices and we discuss their implications for understanding the emergent computations of neural population codes.

Abstract: A central question in neuroscience is how the structure of the brain determines its activity and function. To explore this systematically, we develop large-scale bio-realistic simulations of brain circuits, which integrate a broad array of experimental data: Distribution and morpho-electric properties of different neuron types; Connection probabilities, synaptic weights, axonal delays, and dendritic targeting rules; And a representation of inputs into the simulated circuits from other parts of the brain. We will discuss this approach focusing on the 230,000-neuron model of mouse primary visual cortex (area V1). Simulations of neural activity in the model match experimental recordings in vivo on a number of metrics, such as firing rates, direction selectivity, and others. Applications include the following problems of broad interest: Understanding how architecture of brain circuit gives rise to the observed functional activity; Learning of behavioral and computational tasks in biological and artificial networks; Generation of the extracellular electric potential due to synaptic activity in the cortex. The model is shared freely with the community via brain-map.org, as are the datasets it is based on.

Abstract: This seminar will present the authors' recent preprint on a computational approach to evaluate how molecular mechanisms impact large-scale brain activity.

Topics: Co-Design & Science Support, EBRAINS Base infrastructure and HPC services

Topics: EBRAINS-RI architecture, multiple scales of complexity, EBRAINS Education and events for early career researcher

Place: Zoom (Please find the zoom link by the registration)

Abstract: Factor analysis is a statistical method for identifying latent factors from the correlation structures of high-dimensional data. It was originally developed for applications in social and behavioral sciences but has since been applied to various research fields, including the natural sciences. An advantage of factor analysis is that it leads to interpretable latent factors, enabling applications such as the identification of active brain regions in neuroscience. In this study, we propose a penalized maximum likelihood estimation method aimed at enhancing the interpretability of latent factors. In particular, the Prenet (Product-based elastic net) penalization allows for the estimation of a perfect simple structure, a desirable characteristic in the factor analysis literature. The usefulness of the proposed method is investigated through real data analyses. Finally, we discuss potential extensions and applications of the proposed method in neuroscience.

Location: Okinawa Institute of Science and Technology (on-site only)

Place: Zoom (Please find the zoom link by the registration)

Abstract: わが国は⻑年に渡り数理科学分野で高い水準を維持してきた．しかし現在の社会におけるDX化の流れの中で，産業界・諸科学分野・社会からの課題に応え価値を共創する新たな数理連携基盤を築く必要に迫られている．このような要請に数学コミュニティ全体で応え，総合知構築を実現するオールジャパン体制のプラットフォームの構築を目指して，2023年10月に全国17の数学・数理科学に関連する機関が参画し，マス・フォア・インダストリ・プラットフォーム (MfIP) を発足した．MfIPの活動の一つとして，脳科学分野の皆様と，連携探索ワークショップの開催 (2023年12月28日@日本橋ライフサイエンスビルディング)，Digital Brain Seminar/Workshopの開催 (2024/9/19〜21日@日本橋ライフサイエンスビルディング) などで連携を進めてきた．本チュートリアルでは，さらなる連携の深化を目指して，MfIP의 活動や連携機関に所属する数学者の持つシーズについて紹介する．

Place: Zoom (Please find the zoom link by the registration)

Abstract: Our goal is to discover how the brain produces out emotions, memories, and actions. Answers will be in terms of neural activity in defined neuron types interacting across the whole brain and body. To advance our goals we are developing next-generation neurotechnologies. We are also committed to Open Science: Knowledge, data, and tools will be widely shared, to facilitate science elsewhere and to support the development of therapies for brain disorders. I will describe how AIND scientists and engineers organize for team science. I will then discuss a few recent neurotechnological and scientific advances.

Abstract: How do the diverse functions of the brain originate? Understanding the developmental mechanisms that underlie brain functions is a fundamental question in neuroscience, with significant implications for research on artificial neural networks. This talk will introduce principles related to these developmental mechanisms, which differ notably from the data-driven learning paradigms predominantly used in AI. I will present our recent findings demonstrating that early functional circuits and cognitive functions in the brain can emerge spontaneously, even in the complete absence of training. Using a biologically inspired neural network model, I will first show how regularly structured functional maps can arise from simple local interactions between individual cells. I will discuss how evolutionary variations in physical parameters may lead to the development of distinct functional circuitry in the brain. Next, I will demonstrate that higher cognitive functions, such as visual quantity estimation and primitive object detection, can also emerge spontaneously in untrained neural networks. I will argue that random feedforward connections in early circuits may be sufficient to initiate functional circuits. These findings suggest that early visual functions can emerge from the statistical properties of bottom-up projections in hierarchical neural networks, providing insight into the origins of primitive functions in the brain.

Abstract: This seminar, as part of an ongoing series, will introduce the EBRAINS-RI, featuring a presentation by Prof. Katrin Amunts, co-CEO of EBRAINS. The session will also provide an overview of two key components of the infrastructure: the EBRAINS Knowledge Graph (KG) and the EBRAINS Brain Atlases by Oliver Schmid and Trygve Leergaard respectively. The EBRAINS Knowledge Graph is the metadata management system that underpins EBRAINS' Data and Knowledge Services. It offers essential tools and services designed to make neuroscientific data, models, and software FAIR (Findable, Accessible, Interoperable, and Reusable). EBRAINS Brain Atlases provide comprehensive maps of brain regions defined based on structure, function and neural connections. As spatial reference systems for neuroscience, they are essential for understanding the complexity of the healthy brain, studying brain disorders and seeking to develop new treatments.

Abstract: In the past twenty years, we have made significant progress in creating digital models of an individual’s brain, so called virtual brain twins. By combining brain imaging data with mathematical models, we can predict outcomes more accurately than using each method separately. Our approach has helped us understand normal brain states, their operation and conditions like healthy aging, dementia and epilepsy. Using a combination of computational modeling and dynamical systems analysis we provide a mechanistic description of the formation of resting state manifold via the network connectivity. We demonstrate that the symmetry breaking by the connectivity creates a characteristic flow on the manifold, which produces the major data features across scales and imaging modalities. These include spontaneous high amplitude co-activations, neuronal cascades, spectral cortical gradients, multistability, and characteristic functional connectivity dynamics. When aggregated across cortical hierarchies, these match the profiles from empirical data and explain features of the brain’s microstate organization. The digital brain twin augments the value of empirical data by completing missing data, allowing clinical hypothesis testing and optimizing treatment strategies for the individual patient. Virtual Brain Twins are part of the European infrastructure called EBRAINS, which supports researchers worldwide in digital neuroscience.

Abstract: OptiNiSt (https://optinist.readthedocs.io/) is an open source software tool that helps you to build calcium neuroimage data analysis pipelines by comparing and combining multiple tools through graphical user interface and to produce data processing scripts. In order to promote this tool, developed by the Brain/MINDS project for reproducibility and standardization of neural data analysis, we will hold a hands-on tutorial session. You can use your own note PC to experience how to build a data analysis pipeline. We recommend downloading the Docker image of OptiNiSt according to the instructions (https://github.com/oist/optinist_tutorial_preparation).

Abstract: In this tutorial, we will go over how to use 3D Slicer to load and view NIfTI data, brain atlas and labels. We will be using one ex vivo mouse brain data as an example to perform image registration to a reference template (Turone Mouse Brain Template) space and then apply inverse transforms to the atlas to view it in individual space.

Event: The first Digital Brain Workshop held as an on-site event at Kyushu University Nihonbashi Satellite.

Topic: Tutorial on MRI data usage and analysis for the International Brain Protocol.
Format: Hybrid (Registration necessary).

Abstract: Empirical applications of the free-energy principle at the cellular and synaptic levels are not straightforward because they entail a commitment to a particular process theory (i.e., neuronal basis). To address this issue, we developed a reverse engineering technique that links quantities in neuronal networks to those in Bayesian inference and showed that any canonical neural network—whose activity and plasticity minimise a shared Helmholtz energy—can be cast as performing variational Bayesian inference. By combining with an in vitro causal inference paradigm, we experimentally validated the free-energy principle by showing its ability to predict the quantitative self-organisation of in vitro neural networks. We have recently begun to apply this technique to neural activity of zebrafish and rodents to reverse engineer their generative models. The virtues of the reverse engineering are that, when provided with initial empirical data, it enables the systematic identification of a generative model employed by the biological system. The reconstructed generative model yields a neuromorphic artificial intelligence that performs Bayesian inference. This further enables the quantitative predictions of subsequent self-organisation and learning in the system.

Abstract: Our modern societies strongly rely on machines to survive. These machines treat information in the digital world but interact with humans in the physical world. To bridge the gap between the digital and physical realms, we have dedicated our efforts to the creation of new AI-based 3D vision models. Our research has been centered on the task of capturing and modeling the human body, as this is fundamental for enhancing human-machine interactions. Leveraging generative AI, which learns from vast collections of images and videos, we have been able to gain insights into human body shapes, deformations, and semantic interactions within various scenes. This research represents a significant step toward the development of Large 3D Vision Models, which are essential for advancing machines toward greater autonomy and intelligence. In this talk, I will present the latest innovations in the creation of digital and autonomous human avatars, showcasing how these advancements are shaping the future of human-machine interactions.

Place: Zoom