Dual-Pathway Fusion of EHRs and Knowledge Graphs for Predicting Unseen Drug-Drug Interactions

TL;DR

This work presents a novel dual-pathway framework that fuses knowledge-graph–based DDI reasoning with patient-level EHR context to produce mechanism-specific, auditable alerts. A Fusion Teacher integrates KG relations with EHR-derived drug embeddings, while a Knowledge-distilled Student provides zero-shot, EHR-only predictions for drugs absent from KGs. Across edge and node hold-out evaluations, the approach yields higher precision and F1 than KG-only or unimodal baselines, with case studies demonstrating clinically recognized mechanisms for unseen drugs. The results support practical deployment in clinical decision support and pharmacovigilance, offering interpretable evidence and robust performance across settings. Overall, the combination of KG–EHR fusion and distillation effectively bridges population-level pharmacology knowledge with patient-specific signals, enabling accurate and actionable DDI predictions for both familiar and novel drugs.

Abstract

Drug-drug interactions (DDIs) remain a major source of preventable harm, and many clinically important mechanisms are still unknown. Existing models either rely on pharmacologic knowledge graphs (KGs), which fail on unseen drugs, or on electronic health records (EHRs), which are noisy, temporal, and site-dependent. We introduce, to our knowledge, the first system that conditions KG relation scoring on patient-level EHR context and distills that reasoning into an EHR-only model for zero-shot inference. A fusion "Teacher" learns mechanism-specific relations for drug pairs represented in both sources, while a distilled "Student" generalizes to new or rarely used drugs without KG access at inference. Both operate under a shared ontology (set) of pharmacologic mechanisms (drug relations) to produce interpretable, auditable alerts rather than opaque risk scores. Trained on a multi-institution EHR corpus paired with a curated DrugBank DDI graph, and evaluated using a clinically aligned, decision-focused protocol with leakage-safe negatives that avoid artificially easy pairs, the system maintains precision across multi-institutuion test data, produces mechanism-specific, clinically consistent predictions, reduces false alerts (higher precision) at comparable overall detection performance (F1), and misses fewer true interactions compared to prior methods. Case studies further show zero-shot identification of clinically recognized CYP-mediated and pharmacodynamic mechanisms for drugs absent from the KG, supporting real-world use in clinical decision support and pharmacovigilance.

Figures (1)

• Figure 1: Training pipeline for the distillation model