arxiv_ml 90% Match Research Paper Healthcare Professionals,Medical Informaticians,Pharmacists,AI Researchers in Healthcare,Drug Developers 4 days ago

Traceable Drug Recommendation over Medical Knowledge Graphs

graph-neural-networks › knowledge-graphs

📄 Abstract

Abstract: Drug recommendation (DR) systems aim to support healthcare professionals in selecting appropriate medications based on patients' medical conditions. State-of-the-art approaches utilize deep learning techniques for improving DR, but fall short in providing any insights on the derivation process of recommendations -- a critical limitation in such high-stake applications. We propose TraceDR, a novel DR system operating over a medical knowledge graph (MKG), which ensures access to large-scale and high-quality information. TraceDR simultaneously predicts drug recommendations and related evidence within a multi-task learning framework, enabling traceability of medication recommendations. For covering a more diverse set of diseases and drugs than existing works, we devise a framework for automatically constructing patient health records and release DrugRec, a new large-scale testbed for DR.

Authors (6)

Yu Lin

Zhen Jia

Philipp Christmann

Xu Zhang

Shengdong Du

Tianrui Li

Submitted

October 31, 2025

arXiv Category

cs.IR

arXiv PDF

Key Contributions

This paper introduces TraceDR, a novel drug recommendation (DR) system operating over a medical knowledge graph (MKG) that ensures traceability of recommendations. TraceDR uses multi-task learning to simultaneously predict drug recommendations and related evidence, addressing the critical need for explainability in high-stakes healthcare applications. It also proposes a framework for automatic patient record construction and releases a new large-scale testbed, DrugRec.

Business Value

Improves patient safety and clinical decision-making by providing evidence-based, traceable drug recommendations, potentially reducing medical errors and optimizing treatment outcomes.

Paper Metadata

Innovation Type

System/Algorithmic

Deployment Feasibility

Moderate. Requires integration with existing Electronic Health Record (EHR) systems and access to comprehensive medical knowledge graphs. The multi-task learning approach is standard.

Limitations Addressed

Lack of insights into the derivation process of drug recommendations from deep learning models,Limited scope of existing DR systems (diseases/drugs),Need for traceable and evidence-based recommendations in healthcare

Performance Gains

Traceable drug recommendations,Simultaneous prediction of drugs and evidence,Expanded scope through new testbed

Technical Tags

drug recommendationmedical knowledge graph (MKG)traceabilitymulti-task learningdeep learningevidence predictionpatient recordsdisease predictiondrug interactionrecommendation system

Research Topics

Knowledge GraphsRecommender SystemsMedical InformaticsDrug DiscoveryDeep LearningExplainable AI

Methods & Architectures

Multi-task LearningDeep LearningKnowledge Graph EmbeddingsEvidence Prediction

Applications & Tasks

Healthcare Pharmacology Clinical Decision Support Drug Recommendation Drug RecommendationExplainability in HealthcareKnowledge Graph Reasoning Recommending appropriate medicationsProviding evidence for recommendationsConstructing patient health records automatically

Datasets & Benchmarks

Datasets

DrugRec (new testbed)

Accuracy of drug recommendationsQuality of evidence providedTraceability of recommendations

Related Fields

Knowledge GraphsRecommender SystemsMedical InformaticsExplainable AIDeep Learning

Keywords

drug recommendationknowledge graphmedicaltraceabilitymulti-task learningdeep learningexplainable AIclinical decision supportpatient recordsDrugRec

Academic Context

#Knowledge Graphs#Recommender Systems#Medical Informatics#Drug Discovery#Deep Learning#Explainable AI

Commercial Potential

Potential Products

Clinical decision support systems for medication managementAI-powered pharmaceutical research tools

Target Industries

HealthcarePharmaceuticalsBiotechnologyMedical Devices

Use Case Examples

Assisting doctors in selecting the best drug for a patient's conditionIdentifying potential drug-drug interactionsProviding evidence-based justifications for prescription choices

Competitive Edge

Differentiates itself by providing traceable and evidence-based drug recommendations, a critical feature for high-stakes medical applications, unlike black-box deep learning models.

Resource Requirements

Compute Needs

Moderate to High (training deep learning models on large KGs)

Data Requirements

Medical Knowledge Graph, patient data, drug information

Deployment Constraints

Requires integration with existing healthcare IT infrastructure and data privacy compliance (HIPAA).

Scalability

Scalability depends on the size of the knowledge graph and the efficiency of the multi-task learning framework.

View Full Paper Back to Papers