drGT
Co-developed experiments · Led analysisA heterogeneous graph model connecting drugs, cell lines, and genes to predict response while producing gene-level, mechanism-oriented attributions.
Research portfolio
Selected projects
Methods and resources for predicting therapeutic response, understanding mechanism, and integrating biomedical evidence.
DrugAgent
First author · Framework developmentA multi-agent framework for integrating heterogeneous and conflicting biomedical evidence for drug-target interaction prediction and drug discovery.
Treatment-Conditioned Representation Learning
Lead research directionLearning representations that model biological state together with therapeutic intervention to improve generalization across perturbational and patient contexts.
GraphPINE
First author · Method developmentGraph importance propagation for interpretable drug-response prediction, incorporating prior biological relationships into feature importance learning.
CellMinerCDB 2.2
Collaborative researchA cross-database platform for exploring pharmacogenomic relationships across patient-derived cancer cell-line datasets.
TRPM4 and Acetalax
Collaborative researchInvestigation of TRPM4 as a predictive biomarker and mechanistic driver of Acetalax activity in prostate cancer.
BiGCN
First author · Method developmentA bi-graph convolutional approach leveraging cell and gene similarities for single-cell transcriptome imputation.
Research Toolbox
Open-source developmentReusable utilities for evaluation, reproducibility, molecular fingerprints, biomedical identifier mapping, caching, PMC XML processing, and scientific visualization.
Research Ideas
Public research notebookExploratory questions, candidate projects, and forward-looking directions in therapeutic response modeling and biomedical AI.