01
drGT
Structured representation learning over drug–cell–gene relationships to predict response while exposing biologically meaningful gene-level mechanisms.
Yoshitaka Inoue · UMN / NIH
AI for therapeutic response and precision medicine.
I study how shared biological representations emerge across molecular, cellular, and patient contexts, and how they can support transferable and interpretable models of therapeutic response.
Research
Discovering shared biological structure across contexts
My central question is what biological structure is shared across domains, modalities, and perturbations—and when that structure is transferable, interpretable, and mechanistically meaningful.
Domains & modalities
+
Perturbations
Shared biological
representation
representation
Transfer across contexts
Model intervention
Explain mechanism
Therapeutic response
Selected work
Three parts of the research program
02
DrugAgent
Mechanistic reasoning over heterogeneous and conflicting biomedical evidence, connecting structured knowledge with data-driven drug discovery.
03
Treatment-conditioned representation learning
Current work on latent spaces that preserve transferable biological structure while explicitly modeling intervention and context.
Selected publications
Representative papers
- 2026drGT: Interpretable Drug Response Prediction with Attention-Guided Gene Attribution on a Drug-Cell-Gene Heterogeneous Graph
- 2026TRPM4 Expression as a Predictive Biomarker and a Mechanistic Driver of Acetalax Activity in Prostate Cancer: Preclinical Efficacy Studies
- 2026CellMiner cross-database (CellMinerCDB) version 2.2 for explorations of patient-derived cancer cell line pharmacogenomics
Recent
Health Data Science review accepted.
ISMB 2026 oral presentations for drGT and DrugAgent.
CCR-FYI 2026 Outstanding Postgraduate Fellow finalist.