Supplementary material (Figures) of the paper titled "Restructuring knowledge graphs with conceptual models: implications for machine learning predictions in drug repurposing"

This research explores how restructuring knowledge graphs (KGs) with well-founded conceptual models can improve machine learning (ML) predictions, particularly for drug repurposing. Using OntoUML and the Unified Foundational Ontology, the study applies a FAIRification workflow to enhance data quality. A Graph Neural Network model was trained on both original and restructured KGs, revealing that while classification performance remained similar, the restructured KGs led to more consistent predictions with reduced variability. These findings suggest that conceptual models can enhance the reliability of ML predictions without compromising accuracy, highlighting new directions for future research.本研究探讨了采用具备坚实理论基础的概念模型重构知识图谱（knowledge graphs, KGs）对提升机器学习（machine learning, ML）预测性能的作用，尤其聚焦于药物重定位任务。研究采用本体统一建模语言（OntoUML）与统一基础本体（Unified Foundational Ontology），通过FAIR化工作流提升数据质量。本研究分别基于原始知识图谱与重构后的知识图谱训练图神经网络（Graph Neural Network）模型，结果显示，尽管两类模型的分类性能相近，但基于重构知识图谱得到的预测结果一致性更强、变异程度更低。上述研究结果表明，概念模型可在不降低预测精度的前提下提升机器学习预测的可靠性，为后续研究指明了新的发展方向。