Higher-order drug-drug interaction prediction using Graph Neural Networks. Predicting emergent adverse effects from 3+ drug combinations for safer polypharmacy.
Pairwise drug-drug interactions are well-studied, but patients on polypharmacy often take three or more drugs simultaneously. Emergent interactions — adverse effects that arise only when 3+ drugs are combined — remain largely unexplored computationally.
PolyDDI tackles this problem with a GNN-based pipeline built in PyTorch:
Learns drug embeddings from 191K pairwise DDIs (DrugBank via TDC) over a graph of 1,706 drugs, with finetuning on triplet signal for +0.178 AUROC gain.
Mines emergent 3-way adverse event signals from 8 quarters of FDA FAERS reports (2023Q1–2024Q4) using Proportional Reporting Ratios with FDR correction.
Permutation-validated evaluation with hard negatives (replace 1 drug in positive triplet), giving honest AUROC 0.763 vs. inflated 0.941 with random negatives.
| Setting | Test AUROC | Interpretation |
|---|---|---|
| Finetuned SAGE + Hard Neg | 0.763 | Realistic performance |
| Finetuned SAGE + Random Neg | 0.941 | Upper bound (trivial negatives) |
| Frozen SAGE + Hard Neg | 0.585 | Pairwise embedding alone |
| Random Split + Random Neg | 0.988 | Memorization ceiling |
Cannot predict for drugs absent from the training graph. Inductive molecular encoders (SMILES/atom-level GNN) are the planned next step.
FAERS data contains observation bias, reporting bias, and confounding. Signals are statistical associations, not confirmed causal interactions.
Random negative sampling inflates evaluation metrics. Hard negatives provide a more honest performance estimate.
| Source | Access | Usage |
|---|---|---|
| DrugBank via TDC | Open | 1,706 drugs, 191K pairwise DDIs |
| FDA FAERS | Open | 8 quarters of adverse event reports (2023Q1–2024Q4) |