Sponsored Session
| ▼SS01 | AMED/BINDS | 10/28 14:00-15:30 | [Zuiun] |
| ▼SS02 | GDEP ADVANCE, Inc. | 10/28 14:00-15:30 | [Heian] |
| ▼SS03 | Latent Chemical Space Based on Diverse Natural Products for Bio-active Molecular Design | 10/28 14:00-15:30 | [Fukuju] |
| ▼SS04 | Iktos K.K. | 10/29 14:00-15:30 | [Zuiun] |
| ▼SS05 | Elix, Inc. | 10/29 14:00-15:30 | [Heian] |
| ▼SS06 | GENESIS Users' Group | 10/29 14:00-15:30 | [Fukuju] |
| ▼SS07 | Xeureka Inc. | 10/30 14:00-15:30 | [Zuiun] |
| ▼SS08 | SAKURA internet Inc. | 10/30 14:00-15:30 | [Heian] |
| ▼SS09 | Amazon Web Services Japan G.K. | 10/30 14:00-15:30 | [Fukuju] |
| Sponsored Session SS04 Iktos K.K. |
[Zuiun] 10/29 14:00-15:30 |
| New Frontiers in the Digital Transformation of Chemistry Powered by AI and Robotics | |
Moderator: Hideyoshi FUJI (Iktos K.K.)
| SS04-01 | |
| Ikumi Kuriwaki ( Astellas Pharma Inc.) | |
| "Astellas’s Digital Transformation ～DX-Driven Drug Discovery and DX-Talent Development for Chemical Modalities～" | |
| SS04-02 | |
| Victoire Cachoux（Iktos K.K.） | |
| "Synthesis-driven GenAI for molecule design: Growing and Linking Optimizers " | |
| SS04-03 | |
| Hideyoshi Fuji（Iktos K.K.） | |
| "Iktos' Next-Generation Drug Discovery Approach: 3D Generative AI and Robotics-Driven Laboratory Automation " | |
| Sponsored Session SS05 Elix, Inc. |
[Heian] 10/29 14:00-15:30 |
| Accelerating Drug Discovery with Elix DiscoveryTM: Case Studies and New Features of Our AI Platform for Small-Molecules |
|
Moderator: Asako Tsubouchi (Elix, Inc.)
| SS05-01 | |
| Tomoki Nishioka (Eisai Co., Ltd.) | |
| SS05-02 | |
| Tatsunobu Sumiya (KYORIN Pharmaceutical Co., Ltd.) | |
| SS05-03 | |
| Akimi Hori (Japan Tobacco Inc.) | |
| SS05-04 | |
| Tasuku Ishida (Elix, Inc.) | |
| SS05-05 | |
| Takahiro Inoue (Elix, Inc.) | |
| Sponsored Session SS06 GENESIS Users' Group |
[Fukuju] 10/29 14:00-15:30 |
| 15th GENESIS Users' Group Seminar ~Introduction to molecular dynamics software GENESIS~ | |
Moderator: Suyong Re (National Institutes of Biomedical Innovation, Health and Nutrition (NIBN))
| SS06-01 | |
| Song-Ho Chong (Kumamoto University, Faculty of Life Sciences) | |
| "Elucidating missing structural information in protein systems using gREST enhanced sampling" | |
| How can we characterize the structure of a functionally important protein region when experimental data are lacking, or determine the binding mode of a ligand when structural information is unavailable? While AlphaFold has become a widely used tool, it still faces limitations, particularly in predicting flexible or intrinsically disordered regions and in accurately modeling protein–ligand complexes. In such challenging situations, the generalized replica exchange with solute tempering (gREST) method offers a powerful alternative. gREST is an enhanced sampling approach that selectively applies temperature scaling to user-defined “solute” regions while maintaining the rest of the system at physiological temperature. This strategy enables efficient conformational sampling of structurally elusive regions without disrupting the overall molecular context. In this presentation, we first provide an overview of the gREST methodology and then introduce two of its applications: peptide binding mode prediction for Src tyrosine kinase and structural modeling of the neck region in kinesin motor protein. The first example involves Src tyrosine kinase. Using gREST-based enhanced sampling, we successfully determined the binding mode of a peptide substrate, an experimentally elusive structure that has not been resolved to date. This structural insight provided a critical starting point for subsequent molecular dynamics simulations, which revealed how substrate binding alters the kinase’s conformational landscape and triggers allosteric conformational changes. The second example involves kinesin, a motor protein essential for intracellular transport. Its neck region is known to be critical for directional stepping but has long eluded structural determination. Using gREST, we resolved its atomic-scale structure in the microtubule-bound state. This structural model enabled downstream simulations that revealed how interactions between the neck region and the microtubule play a crucial role in directionally biasing kinesin’s stepping trajectory. Together, these studies highlight the unique power of gREST to uncover functionally relevant structural details and to serve as a foundational tool in dynamic biomolecular modeling. |
|
| SS06-02 | |
| Hiraku Oshima (Graduate School of Science, University of Hyogo) | |
| "Prediction of Protein-Ligand Binding Affinity Using Free Energy Perturbation" | |
| Accurate estimation of protein–ligand binding affinities remains a major challenge in in silico drug discovery, as it directly influences the efficiency and cost-effectiveness of lead optimization. The free energy perturbation (FEP) method, based on all-atom molecular dynamics (MD) simulations, provides a rigorous framework for computing binding and solvation free energies with high accuracy. The FEP implementation in GENESIS [1] supports both relative and absolute binding free energy calculations, as well as relative and absolute solvation free energy estimations, achieving high predictive performance [2,3]. By combining FEP with generalized replica exchange with solute tempering (FEP-gREST), conformational sampling of the protein and ligand is enhanced by selectively increasing the temperature of designated regions in the system. To further improve computational efficiency, especially for large biomolecular systems, we implemented a modified FEP scheme that incorporates non-uniform scaling of Hamiltonian parameters [4]. This modification significantly reduces computational cost while maintaining predictive accuracy. Here we present the technical features of the FEP functions in GENESIS and evaluate their performance in predicting protein–ligand binding affinities. We also discuss future directions for extending the applicability of GENESIS to protein–ligand systems in drug discovery. [1] J. Jung, K. Yagi, C. Tan, H. Oshima, T. Mori, I. Yu, Y. Matsunaga, C. Kobayashi, S. Ito, D. Ugarte La Torre, Y. Sugita, J. Phys. Chem. B 128, 6028-6048 (2024) [2] H. Oshima, S. Re, Y. Sugita, J. Chem. Inf. Model. 60, 5382-5394 (2020) [3] S. Kim, H. Oshima, H. Zhang, N. R. Kern, S. Re, J. Lee, B. Roux, Y. Sugita, W. Jiang, W. Im, J. Chem. Theory Comput. 16, 7207-7218 (2020) [4] H. Oshima, Y. Sugita, J. Chem. Inf. Model. 62, 2846–2856 (2022) |
|
| SS06-03 | |
| Kiyoshi Yagi (Department of Chemistry, Institute of Pure and Applied Sciences, University of Tsukuba) | |
| "Development of QM/MM methods in GENESIS and applications to enzymatic reactions" | |
| Hybrid quantum mechanical / molecular mechanical (QM/MM) method treats a region of chemical interest by highly accurate electronic structure methods while incorporating the surrounding solution and biological environment by low-cost molecular force fields. The QM/MM method has been widely used to elucidate the mechanism of complex systems such as enzyme reactions. We have implemented the QM/MM method into a molecular dynamics (MD) program GENESIS, which calls external QM programs to compute the electronic structure of the QM system. In this talk, I will illustrate the QM/MM module in GENESIS, namely, which QM programs are supported, how the input files are structured, how the parallelization is designed, and what users need to prepare to run the program. I will also show our recent development, which combines GENESIS/SPDYN with QSimulate-QM to achieve high-performance QM/MM-MD simulations [J. Chem. Theory Comput. 2025, 21 4016-4029]. The combination of highly parallelized algorithms in the two programs performs MD simulations based on DFTB in the QM size of ∼100 atoms and MM of ∼100,000 atoms with a better performance than 1 ns/day using one computer node. This feature paves the way for various enhanced sampling simulations based on the QM/MM potential. For example, free-energy perturbation (FEP) calculations based on QM/MM for predicting accurate ligand binding energies to target proteins, and enhanced sampling methods for elucidating the mechanism of enzymatic reactions are under development. | |
| SS06-04 | |
| Yuji Sugita (RIKEN・University of Tokyo) | |
| "Development of multi-scale molecular dynamics software GENESIS: Current Status and Future Plan" | |
| GENESIS is molecular dynamics software developed by RIKEN in collaboration with developers from universities and research institutions across Japan. Its main features include support for multi-scale molecular dynamics (MD), including all-atom MD, coarse-grained MD, and hybrid quantum mechanics/molecular mechanics (QM/MM) MD; efficient structure exploration algorithms, such as replica-exchange MD (REMD) and generalized replica exchange with solute tempering (gREST); and free energy calculations using methods like free-energy perturbation (FEP). GENESIS efficiently performs large-scale MD simulations of biological systems using supercomputers such as “K” and “Fugaku,” and also efficiently performs calculations required for in-silico drug discovery, such as protein-substrate binding affinity and ligand binding pose. In recent years, we have also made significant efforts to improve efficiency in MD simulations with GENESIS on GPU-equipped computers. We have performed advanced optimization not only on Nvidia GPUs but also on AMD GPUs. Using LUMI, a supercomputer equipped with AMD CPU and GPU in Finland, we have demonstrated that all-atom MD simulations of a 16 million-atom system can be completed at a performance exceeding 200 ns/day. We have begun preparations for the release of a new version (2.2.0) of GENESIS that incorporates features already published in papers, including a more efficient coarse-grained MD program, CGDYN, ABMD methods for enhanced sampling with a reduced number of replicas. Our future goals include further optimizing GENESIS on Nvidia and AMD GPUs, as well as creating a Python interface to enable tight integration with machine learning and quantum computing. In the talk, we will also introduce the roadmap for these future developments. | |
| Sponsored Session SS09 Amazon Web Services Japan G.K. |
[Fukuju] 10/30 14:00-15:30 |
| Next-generation drug discovery platform powered by agentic AI | |
Moderator: Yuto Kataoka (Amazon Web Services Japan G.K.)
| SS09-01 | |
| Tetsuto Matsunaga (Amazon Web Services Japan G.K.) | |
| "Next-generation drug discovery platform powered by agentic AI" | |
| SS09-02 | |
| Yuji Shiina (Amazon Web Services Japan G.K.) | |
| "Utilizing AWS services to build infrastructure for genomic analysis and drug discovery research" | |
| SS09-03 | |
| Kentaro Takahashi (Human Biology DataEcosystem, Integrated Data Science, Human Biology Creation Hub, DHBL Eisai Co., Ltd.) | |
| "AI for Human Biology Understanding" | |