计算机科学
药物数据库
人工智能
水准点(测量)
机器学习
超图
一致性(知识库)
图形
构造(python库)
地点
人工神经网络
二部图
鉴定(生物学)
光学(聚焦)
特征学习
深度学习
理论计算机科学
语义属性
先验与后验
交互网络
数据集成
代表(政治)
匹配(统计)
作者
Lan Huang,Zhiyao Han,Jiana Ding,Zhen Jing,Nan Sheng,Yan Wang
标识
DOI:10.1021/acs.jcim.6c01669
摘要
Abstract Drug-food interactions (DFIs) are one of the major factors that trigger adverse drug events and have important research and clinical application value. However, experimental identification of DFIs remains costly and limited in scalability, leading to insufficient coverage of known interactions. Existing computational methods often treat DFI prediction as a variant of drug-drug interaction (DDI) prediction and primarily focus on drug-food constituent relationships, which restrict their ability to directly model drug-food interactions. In this study, we propose a cross-view hypergraph-graph contrastive learning framework named HyperDFI for predicting interactions between drug and food (food constituents). We first construct four benchmark data sets by integrating information from DrugBank and PubMed, covering both drug-food and drug-food constituent interactions. Based on these data sets, heterogeneous bipartite graphs were established. HyperDFI jointly captures local topological structures and global high-order semantic relationships by combining graph neural networks (GNNs) and dynamic hypergraph neural networks (HGNNs). Furthermore, a cross-view contrastive learning strategy is introduced to align representations at different structural scales, thereby enhancing their consistency and discriminability. Extensive experiments on four benchmark data sets demonstrate that HyperDFI consistently outperforms state-of-the-art methods. In addition, case studies on grapefruit further validate its ability to uncover novel DFIs.
科研通智能强力驱动
Strongly Powered by AbleSci AI