Adhikari, D.; et al. "Uncertainty-Aware Drug-Target Affinity Prediction with Budget-Constrained Compound Ranking." IEEE International Conference on Biomedical and Health Informatics (BHI), 2026.

This project develops a graph-neural-network pipeline for drug-target binding prediction with uncertainty estimates that let downstream ranking abstain on low-confidence compounds. A GIN ensemble achieved CI=0.882 on KIBA, CI=0.735 on Davis, and CI=0.625 with AUROC=0.760 on BindingDB.

I also built a multi-agent RAG architecture for biomedical reasoning and decision support. The reproducibility package includes code, experiments, a LaTeX paper, a poster, and training logs. This was built for the GT STAR-AI Makerspace Hackathon on May 12, 2026, at the Parker H. Petit Institute for Bioengineering in Atlanta, crossing from my formal thermal-systems background into graph ML, uncertainty quantification, and bioinformatics.