Topics

Presentation on Prediction Model for Brain Concentration-Time Profiles of Compounds by Dr. Handa (Digital Unit) at the Structure-Activity Forum 2025

2025.06.10

Dr. Handa from the Digital Unit will present a prediction model for brain concentration–time profiles of compounds at the Structure–Activity Forum 2025, organized by the Structure–Activity Relationship Division, The Pharmaceutical Society of Japan, to be held online on June 20, 2025.
If you are planning to attend the forum, we encourage you to join the presentation.

Conference Name: Structure–Activity Forum 2025
Organized by: The Division of Structure–Activity Studies , The Pharmaceutical Society Japan
Co-sponsored by: Chem-Bio Informatics Society (CBI Society), Japan

Date & Time: Friday, June 20th, 10:40 – 11:20 am (JST)

Location: Online

Title: Advancing Practical Machine Learning Models in Drug Discovery

Presenter: Koichi Handa

Presentation Summary
With recent advances in AI technology, the importance of predictive models that leverage large-scale, high-quality data has grown significantly in the field of drug discovery. Our company has developed predictive models capable of accurately estimating 23 pharmacokinetic and toxicity-related endpoints—such as solubility, membrane permeability, and toxicity—based on in vitro and in vivo data collected under standardized conditions. In particular, predicting brain penetration is critical for central nervous system (CNS) disorders, but conventional methods have limitations. To address this, we have developed a Deep two-compartment model that predicts brain concentration-time profiles. This model optimizes compartment parameters using actual measurement data, enabling more practical and reliable predictions. Through the continuous development and application of our AI platform, we aim to address these challenges. In this presentation, we will share insights gained from this ongoing work.

Axcelead DDP’s Soulution
At Axcelead DDP, we are advancing AI development through close collaboration between data scientists proficient in cutting-edge machine learning technologies and researchers with extensive knowledge and experience in drug discovery. Our goal is to further improve the efficiency and success rate of drug development. Axcelead DDP’s AI is characterized by the following features:
1. Utilization of high-quality, large-scale drug discovery data accumulated over many years.
2. Adoption of foundation models using advanced computational algorithms such as graph neural networks and Transformers, which enable more complex and sophisticated analysis.
3. Implementation of user-friendly features, including estimation of the model’s applicability domain and prediction error, making the system more reliable and practical for real-world use.
We currently offer ADME prediction models that greatly enhance the efficiency of compound optimization. Moving forward, we will expand our AI model development to include predictions for compound activity and toxicity. If you are considering AI-driven drug discovery, please feel free to consult with us.

 

Koichi Handa holds a Ph.D. from Kitasato University and completed graduate studies at the University of Tokyo. He has extensive experience in the pharmaceutical industry, having worked at Toyama Chemical, Fujifilm, and Teijin Pharma. He also served as a visiting researcher at the University of Cambridge. Currently, he works at Axcelead Tokyo West Partners. His expertise lies in Drug Metabolism and Pharmacokinetics (DMPK), with a strong focus on applying machine learning to pharmacokinetic modeling and drug development. He had served as the project leader in AI consortium LINC, also serves as the representative of the AI/ML group in the DIS of JSSX, integrating experimental and computational approaches to optimize drug discovery.

Masaki Saito received his Ph.D. in Information Science from the Graduate School of Information Sciences at Tohoku University in March 2016. He earned his doctorate in Information Science. From April 2016, he has been engaged in applied research on machine learning, primarily focusing on computer vision, at Preferred Networks, Inc. Starting in September 2024, he will conduct research and development on drug discovery using machine learning at Axcelead Drug Discovery Partners.

Dr. Hiroshi Kajino is a researcher on machine learning and its application to chemistry. He obtained his Ph.D. degree in 2016 from the University of Tokyo on privacy-preserving techniques for machine learning, and since then he had been a research staff member at IBM Research – Tokyo until February 2024. During his tenure in IBM Research, not only has he regularly published research papers in top-tier conferences but also has contributed to business applications of machine learning to a wide variety of industry, including finance and materials science. He also led several research projects in IBM Research including machine learning for drug discovery. His current research interest is optimization of molecular structures using machine learning. He has published a Japanese monograph on the research topic and has co-authored a Japanese book on reinforcement learning.