Wataru Miyahara
Researcher dedicated to contributing to drug discovery and toxicology through Interpretable Machine Learning.
Meiji Pharmaceutical University, Faculty of Pharmacy (Class of 2028)
Research Affiliation: Laboratory of Medical Molecule Analysis (Members)
Research Philosophy
- ML for Scientific Discovery:
Beyond just predictive accuracy, I focus on Interpretable ML—techniques that present "why a model made a prediction" in chemically and biologically meaningful ways. My goal is to ensure AI acts as an exploratory tool for researchers rather than a black box, ultimately advancing drug discovery processes and safety assessments.
- Key Methodologies:
Current work includes identifying Toxicophores using Kolmogorov-Arnold Networks (KAN) interpretability, optimizing structural descriptors, and multimodal analysis using Graph Neural Networks like GINE-Net.
My Journey (Recent Achievements)
→ View all achievements
Skills & Experience
- Machine Learning (Interpretable ML / KAN / GNN): Architecture design focused on model interpretability.
- CompChem: Feature extraction via RDKit and application of MOPAC/xTB workflows.
- Software Development: Developer of ecfp_cli and mordred_descriptor_calculator, used for data preprocessing.
→ View project details
Contact & Profiles