Application of knowledge graphs in rare disease research

TL;DR

This review discusses how knowledge graphs can help overcome data challenges in rare disease research by integrating diverse data and improving diagnosis and treatment.

Contribution

The paper highlights novel integrations of knowledge graphs with large language models for better medical decision-making in rare diseases.

Findings

KGs help integrate multimodal data using standardized ontologies like HPO.

KGs improve diagnosis and drug repositioning through semantic reasoning and graph neural networks.

Combining KGs with LLMs enhances interpretability and precision in rare disease research.

Abstract

Rare disease research faces significant challenges due to data sparsity and heterogeneity, leading to diagnostic delays and limited treatments. Knowledge Graphs (KGs) offer a computational solution by integrating multimodal data into structured semantic networks. This review explores the technical paradigms and applications of KGs throughout the rare disease workflow. We first describe the data foundation, focusing on standardized ontologies (e.g., HPO) and integration strategies. Subsequently, we examine core applications in elucidating pathogenic mechanisms via link prediction, enhancing clinical diagnosis through semantic reasoning, and optimizing drug repositioning using Graph Neural Networks. Notably, the review highlights the emerging integration of KGs with Large Language Models (LLMs), particularly Retrieval-Augmented Generation (RAG), to improve interpretability and precision in medical decision-making. Finally, we discuss challenges such as privacy and dynamic updates, proposing future directions like federated learning to advance the field.

Figures