Selected in recognition of the presentation titled 'Construction of a Risk Classification Prediction Model for Drug-Induced Liver Injury and Analysis of Associated Chemical Structural Features Using FAERS.'
Awarded for the presentation titled 'Elucidation of structural features and development of predictive models for rhabdomyolysis using FAERS and Kolmogorov-Arnold Networks.'
Multimodal Strategy: Integrated pre-calculated quantum chemistry results (e.g., MACE-xTB) and 1,800+ descriptors. Specifically, developed a specialized feature space for fluorescence prediction by introducing unique Conjugation Features to evaluate the quality of π-conjugated systems.
Wataru Miyahara, Yoshihiro Uesawa*. "Interpretable Toxicity Prediction Using Fragment–Molecular Evidence Stacking: Explicit Integration of Local Fragment and Whole-Molecule Evidence". Preprints.org (Preprint), 2026-06-29.
[Paper/Preprint]
Introduces a novel stacking methodology integrating local structural fragment features with whole-molecule descriptors for highly interpretable toxicological machine learning models.
Preprint detailing the multimodal quantum chemistry feature integration and Sequential Stacking architecture used to win the EUOS25 challenge fluorescence track.
Presentations
Wataru Miyahara, Yoshihiro Uesawa. "Evaluation of the Utility of Kolmogorov-Arnold Networks in Predicting Drug-Induced Adverse Event Risks". The 53rd Annual Meeting of the Japanese Society of Toxicology, Poster, 2026-07-01.
Wataru Miyahara, Yoshihiro Uesawa. "Construction of a Risk Classification Prediction Model for Drug-Induced Liver Injury and Analysis of Associated Chemical Structural Features Using FAERS". The 9th JSPHCS Freshers Conference, Poster, 2026-07-04.
Wataru Miyahara, Mizuho Asada, Yoshihiro Uesawa. "Elucidation of structural features and development of predictive models for rhabdomyolysis using FAERS and Kolmogorov-Arnold Networks.". The 146th Annual Meeting of the Pharmaceutical Society of Japan, Poster (27-51-pm1-04S), 2026-03-27.