Jikkenigaku Special Issue 'AI and Data-Driven Drug Discovery Research' Published — CTO Tsuzuki Authors Chapter on AI-Driven Drug Target Discovery

Yodosha has published the specialized book Jikkenigaku Special Issue Vol.44 No.5: AI and Data-Driven Drug Discovery Research — Finding More Reliable Therapeutic Targets and Designing Drugs through Multi-omics x Chemoinformatics. Our CTO Taku Tsuzuki authored Chapter 1, Section 6: “Breaking Through the ‘Cognitive Bottleneck’ with AI-Driven Knowledge Mining — Autonomous Exploration of ‘Next Moves’ in Individual Molecular Analysis.” This contribution examines a framework for enhancing decision-making in the early stages of target discovery using AI, increasing the probability of reaching novel target candidates and improving exploration productivity under limited experimental resources.

Book Overview

In recent years, the use of AI and large-scale data has been fundamentally transforming drug discovery research. This book provides a systematic overview of the current state of AI and data-driven drug discovery, spanning both multi-omics and chemoinformatics perspectives, covering drug target exploration, drug candidate molecule discovery and optimization, and elucidation and evaluation of mechanisms of action.

Each chapter features leading researchers explaining specific approaches and practical knowledge, with coverage that extends beyond research methods to include major domestic projects and corporate case studies. This book offers multifaceted insights for researchers and practitioners seeking to understand how to implement AI into existing drug discovery workflows to identify targets and compounds more efficiently and reliably.

Our Contribution

Our contribution addresses the “sequential molecular selection problem” — determining the order in which genes associated with disease phenotypes should be validated. In current life sciences, while the volume of obtainable data has increased dramatically, the step of properly interpreting it and deciding on the next experimental plan remains heavily dependent on human cognitive capacity. As a result, exploration targets tend to be biased toward known or trending molecules, and it has been repeatedly noted that many molecules remain uninvestigated.

This chapter attempts to solve this challenge through a swarm search algorithm that mimics the behavior of the human scientific community. Specifically, multiple AI agents with different exploration policies independently score candidates using inputs such as literature, omics data, and transcriptional regulatory network structures, sharing experimental results across all agents and reflecting them in subsequent candidate selection. By maintaining diverse hypothesis formation while dynamically adjusting agent influence based on results, the approach balances deep exploration of known domains with pioneering of unexplored areas, enabling rational generation of experimental plans.

In a demonstration targeting cellular senescence, 5 out of 8 previously unreported candidate genes extracted from transcriptional regulatory networks significantly suppressed expression of the senescence marker CDKN2A. This approach contributes to improving productivity in the upstream stages of drug discovery by directing limited experimental resources toward more promising novel target candidates while suppressing bias toward known molecules.

Book Information

• Title: Jikkenigaku Special Issue Vol.44 No.5 “AI and Data-Driven Drug Discovery Research — Finding More Reliable Therapeutic Targets and Designing Drugs through Multi-omics x Chemoinformatics”
• Editors: Katsuyuki Yunoki, Yoshihiro Yamanishi
• Publisher: Yodosha
• Release Date: March 5, 2026
• ISBN: 978-4-7581-0433-3
• Price: 6,160 yen (tax included)