Focused Session
| FS01 | Founding Meeting of the Discussion Group on AI-assisted Drug Design and Mesoscopic Chemistry | 10/28 14:00-15:30 | [Togen] |
| FS02 | Meet the Legend | 10/28 14:00-15:30 | [Training room] |
| FS03 | Advances in Foundation Models: Transforming Drug Discovery and Life Sciences | 10/28 14:00-15:30 | [401] |
| FS04 | Advanced Measurement and Analysis | 10/28 13:00-16:00 | [407] |
| FS05 | Cohort x Real-World Data Research Forum | 10/28 14:00-15:30 | [406] |
| FS06 | Medical Data AI Analysis Practical Forum | 10/29 14:00-15:30 | [Togen] |
| FS07 | Frontiers in In Silico Approaches to Pharmacokinetics and Toxicity Prediction: From PBPK Modeling to Target Identification | 10/29 14:00-15:30 | [Training room] |
| FS08 | Founding Meeting of the Discussion Group on AI-assisted Drug Design and Mesoscopic Chemistry (2) | 10/29 14:00-15:30 | [401] |
| FS09 | Molecular Robotics Research Group : Establishing a foundation for systems behavior design in molecular robotics | 10/29 14:00-15:30 | [407] |
| FS10 | How to Build a Career as a Computational Scientist: Learning from Diverse Career Paths | 10/29 14:00-15:30 | [406] |
| FS11 | Toxicity Prediction Powered by Computational Science: The Stem Cell × AI Challenge | 10/30 14:00-15:30 | [Training room] |
| FS12 | Founding Meeting of the Discussion Group on AI-assisted Drug Design and Mesoscopic Chemistry (3) | 10/30 14:00-15:30 | [407] |
| FS13 | Workshop on the Omics Principle | 10/30 14:00-15:30 | [406] |
| FS01 | [Togen] 10/28 14:00-15:30 |
| Founding Meeting of the Discussion Group on AI-assisted Drug Design and Mesoscopic Chemistry | |
Moderator: Kohtaro Yuta (In silico data, Ltd.), Kazuyoshi Ikeda (Institute of Physical and Chemical Research)
| FS01-01 | |
| Kohtaro Yuta (In silico data, Ltd.) | |
| "Report on the Establishment of the 'Discussion Group on AI-assisted Drug Design and Mesoscopic Chemistry': Development of an Overview of AI-assisted Drug Design - Data Science Approaches, Deep Learning, and Generative AI" |
|
| FS01-02 | |
| Tsuneaki Sakata (Co-Creation Bureau, the University of Osaka) | |
| FS01-03 | |
| Kazuyoshi Ikeda (Institute of Physical and Chemical Research) | |
| FS02 | [Training room] 10/28 14:00-15:30 |
| Meet the Legend | |
Moderator: Hirofumi Watanabe (WithMetis Co., Ltd.), Natsumi Miyano (Teijin Pharma Limited)
| FS02-01 | |
| Midori Kamimura (CBI Research Institute Quantum-Structural Life Science Laboratories) | |
| "Integrated Structural Biology in Drug Discovery" | |
| FS03 | [401] 10/28 14:00-15:30 |
| Advances in Foundation Models: Transforming Drug Discovery and Life Sciences | |
Moderator: Fumihito Ushiyama (SyntheticGestalt KK)
| FS03-01 | |
| Norikazu Saiki (Institute of Science Tokyo) | |
| "Engineering super-complex multicellular tissues powered by single-cell foundation model" | |
| FS03-02 | |
| Nicholls Joel (SyntheticGestalt KK) | |
| "10 Billion Compound Pretrained Foundational Model for Drug Discovery" | |
| The process of drug discovery remains extremely costly, time-consuming and prone to high attrition rates. A substantial proportion of clinical trial failures (estimated at 40-50%) [1] stem from poor safety profiles, or lack of efficacy, and this indicates certain fundamental challenges in the early stages of discovery, including target validation and the identification of truly effective lead compounds. Predicting molecular properties in early stages enables prioritization of candidate compounds which have a higher chance of success in clinical trials. Addressing this challenge, we introduce a novel foundational large molecular model pre-trained on an extensive dataset of approximately 10 billion chemical conformers. The model transforms compounds to an embedding which can be used for a broad range of property prediction tasks. Our methodology leverages Pseudo Multi-Parameter Persistent Homology (Pseudo-MPPH) to generate rich topological descriptors from 3D molecular conformations. The Pseudo-MPPH features are stochastically masked in pre-training and fed to a specialized transformer-based architecture, which learns to reconstruct the original unmasked Pseudo-MPPH features by using a self-supervised reconstruction loss. Through training on 10 billion diverse chemical conformers, the underlying structural syntax governing molecular structure is learned. The wide variety of data in this process allows our foundational model to develop rich, generalizable representations that capture fundamental chemical principles, moving beyond simple correlations within specific datasets. [1] Sun D, Gao W, Hu H, Zhou S. Why 90% of clinical drug development fails and how to improve it? Acta Pharm Sin B. 2022 Jul;12(7):3049-3062. doi: 10.1016/j.apsb.2022.02.002. Epub 2022 Feb 11. PMID: 35865092; PMCID: PMC9293739 |
|
| FS03-03 | |
| Junya Yamagishi (Preferred Networks, Inc.) | |
| "Applications of Universal Neural Network Potential in Drug Discovery" | |
| FS03-04 | |
| Yoh Terada (Alivexis, Inc.) | |
| "The concept of the next gen. drug discovery AI, using scaled and high-quality synthetic data produced by ModBind simulation" | |
| FS04 | [407] 10/28 13:00-16:00 |
| Advanced Measurement and Analysis | |
| Moderator: | Hisashi Tadakuma (ShanghaiTech University), Seiichi Ishida (National Institute of Health Sciences/ Department of Applied Life Science, Sojo University), Satoshi Fujita (Photo BIO-OIL, AIST) | |
| FS04-01 | ||
| [13:00-13:30] Shintaro Iwasaki (RIKEN Cluster for Pioneering Research / Graduate School of Frontier Sciences, The University of Tokyo) |
||
| "Advanced tricks in ribosome profiling" | ||
| FS04-02 | ||
| [13:30-14:00] Ryuji Igarashi (School of Life Science and Technology, Institute of Science Tokyo / National Institutes for Quantum Science and Technology (QST)) |
||
| "Pushing the Limits of Quantitative Biosensing with Nanoscale Quantum Sensors" | ||
| FS04-03 | ||
| [14:00-14:30] Sayaka Deguchi (Institute of Integrated Research, Institute of Science Tokyo) |
||
| "Investigating intestinal function and disease pathophysiology using iPS cell and MPS technologies" | ||
| FS04-04 | ||
| [14:30-15:00] Kenji Kawabata (National Institutes of Biomedical Innovation, Health and Nutrition) |
||
| "Improved human blood-brain barrier model for microphysiological system" | ||
| FS04-05 | ||
| [15:00-15:30] Daiju Yamazaki (National Institute of Health Sciences) |
||
| "Image analysis of cardiac toxicity (tentative)" | ||
| FS04-06 | ||
| [15:30-16:00] |
||
| FS06 | [Togen] 10/29 14:00-15:30 |
| Medical Data AI Analysis Practical Forum | |
Moderator: Satoshi Mizuno (Tohoku University), Ryosuke Kojima (Kyoto University), Soichi Ogishima (Tohoku University)
| FS06-01 | |
| Kenta Matsumura (Aomori University of Health and Welfare) | |
| "Real-World-Oriented Machine Learning Analysis in Social Medicine, Clinical Medicine, and Digital Health" | |
| FS06-02 | |
| Akira Izumi (RyuWell Co., Ltd/University of the Ryukyus) | |
| "Social Implementation of AI-Based Dementia Diagnosis: Real-World Validation Using Wearables, Video, and Audio Data" | |
| FS07 | [Training room] 10/29 14:00-15:30 |
| Computational ADMET research group Frontiers in In Silico Approaches to Pharmacokinetics and Toxicity Prediction: From PBPK Modeling to Target Identification |
|
Moderator: Yoshihiro Uesawa (Meiji Pharmaceutical University), Kiyoshi Hasegawa (TechnoPro R&D)
| FS07-01 | |
| Kiyoshi Hasegawa (TechnoPro R&D) | |
| "New Molecular Design Using PBPK Simulations and Machine Learning Models" | |
| FS07-02 | |
| Koichi Handa (Meiji Seika Pharma Co., Ltd.) | |
| "Research Trends in Prediction Models for Physicochemical Properties and In Vitro ADME Parameters" | |
| FS07-03 | |
| Kouichi Yoshinari (School of Pharmaceutical Sciences, University of Shizuoka) | |
| "Development of Toxicity Prediction Methods Using Read-Across with Chemical Structure and Biological Activity Information" | |
| FS07-04 | |
| Kazuki Takeda (Laboratory of Toxicology, School of Veterinary Medicine, Kitasato University) | |
| "Binding Proteomics: A Comprehensive Molecular Docking–Based Target Discovery Method for the Structural Proteome" | |
| Moderator: | Kohtaro Yuta (In silico data, Ltd.), Kazuyoshi Ikeda (Institute of Physical and Chemical Research), Kazufumi Ohkawa (SHIONOGI & CO., LTD.) | |
| FS08-01 | ||
| Yukio Tada (Chem-Bio Informatics Society (CBI)) | ||
| FS08-02 | ||
| Takeru Kameda (SHIONOGI & CO., LTD.) | ||
| "Integration of Deep Learning and Computational Science: Prospects for Drug Discovery Using Boltz-2 and Physical Chemistry Calculations" | ||
| FS08-03 | ||
| FS09 | [407] 10/29 14:00-15:30 |
| Molecular Robotics Research Group | |
Moderator: Ken Komiya (Japan Agency for Marine-Earth Science and Technology (JAMSTEC))
| FS09-01 | |
| Yusuke Himeoka (Universal Biology Institute, Graduate School of Science, The University of Tokyo) | |
| "A theory of cell death and the rarity of stably functioning biochemical systems" | |
| Understanding the inherent difference between life and non-life is one of the central topics in biology. For dissecting life and non-life, development of the theory of cell death and that of difficulty of constructing life-like system would be indispensable. In this talk, I will present our recent attempt on these topics. Currently, the criteria for microbial cell death are purely experimental. There is debate about whether different assays lead to inconsistent judgements of cell viability. Therefore, it would be beneficial for the field of biology to define 'death' mathematically. However, attempts to develop theoretical frameworks for cell death remain largely unexplored. In the present project, our aim was to develop a cell death framework based on controllable cellular states. We start by defining 'dead states' as cellular states that cannot be returned to 'representative living states', regardless of controllable parameters such as gene expression levels and external culture conditions. The representative living state is a reference state of ‘living’ and an intracellular state of steadily growing cells is a straightforward choice for microbes. This definition requires a method to compute restricted, global and nonlinear controllability, for which there is currently no general theory. We have developed ‘Stoichiometric Rays', a simple method for solving controllability computations in catalytic reaction systems. This enables us to determine how enzyme concentrations should be modulated to transition a metabolic state from one state to another. We have used the stoichiometric rays to compute the controllability, and therefore the dead states, of a kinetic model of E. coli central carbon metabolism. We also quantified the boundary dividing the phase space into live and dead states, termed the 'Separating Alive and Non-life Zone (SANZ) hypersurface'. Another approach to analysing life-death boundaries is to quantify the rarity of biochemical systems performing certain functions, such as cell growth and ATP production. We have developed a method to compute the steady states of a kinetic model of cellular metabolism with ultra-high computational efficiency. Using the steady-state enumeration method, we quantified the rarity of homeostatically functioning steady metabolic states. In the talk, I will first present our theoretical framework of cell death and then discuss our preliminary findings on the quantification of rarity. |
|
| FS09-02 | |
| Takashi Nakakuki (Department of Intelligent and Control Systems, Faculty of Computer Science and Systems Engineering, Kyushu Institute of Technology) | |
| "Programmable Biomolecular Circuits for Chemical AI: Toward Memory and Learning Functions" | |
| In this presentation, we introduce a design and implementation strategy for biomolecular circuits that possess memory and learning capabilities, aiming to realize adaptive signal processing for a chemical AI system. This research was performed as a part of the Molecular Cybernetics Project in Grant-in-Aid for Transformative Research Areas (A), which seeks to develop intelligent molecular systems. Recent advances in dynamic DNA nanotechnology have demonstrated the feasibility of DNA-based computation. However, realizing plasticity—the capability to adapt based on past inputs—remains a major challenge. In the project, we mainly addressed three key themes essential for chemical AI systems: (1) Resetting for repetitive inputs, (2) Memory function, and (3) Learning function such as acquisition of conditional reflex. For Theme 1, we tackled the problem of once-only responses by designing circuits that can be reset and reused. We explain two approaches: (i) irradiation-based resetting, using light-sensitive azobenzene-modified DNA to switch states reversibly; and (ii) degradation-based resetting, which uses enzymatic degradation of DNA/RNA strands to reset the system. Both approaches were experimentally validated with successful repetitive input-output cycles, proving their applicability in wet-lab settings. In Theme 2, we developed a molecular memory module (MMM) capable of storing external input information. The MMM was designed as a DNA reaction network with high modularity and theoretical robustness. We applied singular perturbation theory to reformulate its dynamics into a two-timescale system, enabling robust signal retention during memory readout. Simulation and circuit integration tests confirmed that the module could accumulate and preserve molecular information as intended. Theme 3 focuses on constructing a plastic DNA circuit capable of acquiring a conditional reflex, analogous to the transition from a YES gate to an OR gate upon training. This was achieved by integrating the resetting, memory, and update mechanisms into a single DNA circuit. We demonstrated this adaptive behavior through simulation and experimental validation, using input-triggered molecular updates and enzymatic degradation. Further generalization to a multi-input system was also explored, showing scalability and modular design viability. Overall, we succeeded in building DNA-based signal processing with adaptive behavior. Our design principles will provide a theoretical and experimental foundation for future developments in chemical AI systems, potentially enabling molecular devices that respond dynamically to environmental stimuli and learn from them over time. |
|
| FS09-03 | |
| Hafumi Nishi (Graduate School of Information Sciences, Tohoku University; Faculty of Core Research, Ochanomizu University) | |
| "Exploring Protein Sequence Space with Generative AI: Embeddings, Generation, and Interpretability" | |
| Recent advances in protein-related AI technologies, including AlphaFold for structure prediction and generative approaches such as ESM3, have significantly impacted how researchers approach protein analysis and design. These technologies now enable the generation of novel protein sequences, potentially exploring areas of protein sequence space not represented by naturally occurring proteins. However, practical application remains challenging, with generated sequences often diverging structurally and functionally from known protein families. In this talk, I will first briefly survey recent developments in protein generative AI, which have demonstrated varying degrees of success in navigating protein sequence and structure spaces. In our ongoing project, we aim to generate protein sequences that can bridge gaps between existing protein families by controlling the generative process. We leverage embedding representations from models like ESM2 to guide sequence generation. Preliminary results indicate that unguided generation often leads to sequences far from known functional regions, confirmed by structural assessments using AlphaFold2. One critical challenge is interpreting and visualizing these high-dimensional protein embeddings. Our recent experiments, which include UMAP visualizations of generated versus natural protein sequences, highlight the difficulty in understanding what these embeddings represent biologically. We explore various approaches, such as targeted "inpainting" of protein segments, to enhance both the quality and interpretability of generative outputs. Finally, I will share detailed insights and preliminary findings from our practical experiments with generative protein AI, highlighting both its promising applications and current limitations. |
|
| FS10 | [406] 10/29 14:00-15:30 |
| How to Build a Career as a Computational Scientist: Learning from Diverse Career Paths | |
| Moderator: | Kan Shiraishi (Daiichi Sankyo Co., Ltd.), Yusuke Tateishi (Graduate School of Science and Technology, Kumamoto University), Hirofumi Watanabe (WithMetis Co., Ltd.), Keiko Kumazawa (Teijin Pharma Ltd.) | |
| FS10-01 | ||
| Akitoshi Okada (Daiichi Sankyo Co., Ltd.) | ||
| "In midst of converging my career" | ||
| FS10-02 | ||
| Mari N. Itoh (Compound Library Screening Center, Graduate School of Pharmaceutical Sciences, The University of Osaka) | ||
| "A research question and a wonderful encounter led me to research" | ||
| FS10-03 | ||
| Shinichi Katakura (Japan Biological Informatics Consortium) | ||
| "What a former pharmaceutical company researcher thinks about academic-industrial collaboration" | ||
| FS11 | [Training room] 10/30 14:00-15:30 |
| Toxicity Prediction Powered by Computational Science: The Stem Cell × AI Challenge | |
Moderator: Hideko Sone (Yokohama University of Pharmacy)
| FS11-01 | |
| Fumiaki Nakamura (UssioBIO) | |
| "Safety Evaluation of Cosmetic Ingredients Using StemPanTox Alpha" | |
| FS11-02 | |
| Tsuyoshi Kato (Faculty of Informatics, Gunma University) | |
| "Comparative Evaluation of Machine Learning Models Implemented in StemPanTox Beta: Focus on Developmental Data" | |
| FS11-03 | |
| Yoichi Nakao (Graduate School of Advanced Science and Engineering, Waseda University) | |
| "Application of StemPanTox Alpha: Predicting the Biological Activity of Natural Chemical Compounds" | |
| FS012 | [407] 10/30 14:00-15:30 |
| Founding Meeting of the Discussion Group on AI-assisted Drug Design and Mesoscopic Chemistry (3): Discussion on AI and IT-related Technologies in AI-assisted Drug Design | |
| Moderator: | Kohtaro Yuta (In silico data, Ltd.), Takahiro Ikusima (Mathematical Advanced Technology Research Institute Co., Ltd.) | |
| FS12-01 | ||
| Takahiro Ikusima (Mathematical Advanced Technology Research Institute Co., Ltd.) | ||
| "What is drug discovery? We will discuss drug discovery in a broad sense. We will also discuss various mesoscopic challenges, aspects of life management systems, and cutting edge AI tools." | ||
| FS12-02 | ||
| Shinya Yuki (Elix, Inc.) | ||
| FS12-03 | ||
| Masaki Iwatani (NVIDIA Corporation) | ||