| 引用本文: | 杨静,罗音,沈笑,周犀.基于异构图注意力网络的水稻[]代谢物和蛋白质关联预测[J].生物信息学,2026,24(3):240-252. |
| Yang Jing,Luo Yin,Shen Xiao,Zhou Xi.Prediction of rice metabolite-protein interactions based on heterogeneous graph attention network[J].Chinese Journal of Bioinformatics,2026,24(3):240-252. |
|
| 摘要: |
| 在调节蛋白质功能和控制各种细胞过程中,代谢物与蛋白质的相互作用(Metabolite-protein interaction, MPI)至关重要。针对利用实验方法和高通量代谢组和转录组联合分析的传统检测MPI的方法,导致耗费物力和时间成本高的问题,提出了将深度学习领域新兴的异构图注意力网络应用于植物MPI预测。因此,为了预测水稻的MPI,提出建立水稻代谢物与蛋白质关联预测模型OryHNCGAT(Oryza heterogeneous neighbor contrastive graph attention network ),OryHNCGAT模型通过引入注意力机制来聚合邻居层面的信息和关系层面的信息,得到一个较为全面的节点表示。其次添加异构邻居对比学习,将学习到的节点嵌入能够很好地保存异构网络拓扑。实验结果显示,OryHNCGAT模型实现了有效的水稻代谢物与蛋白质关联预测,为进一步研究代谢物与蛋白质相互作用提供了新的方法和途径。 |
| 关键词: 神经网络 异构图注意力网络 异构邻居对比学习 水稻 代谢物与蛋白质相互作用 |
| DOI：10.12113/202407008 |
| 分类号:TP183 |
| 文献标识码:A |
| 基金项目:海南省自然科学基金高层次人才项目(No.322RC570). |
|
| Prediction of rice metabolite-protein interactions based on heterogeneous graph attention network |
|
Yang Jing1,Luo Yin1,Shen Xiao2,Zhou Xi1
|
|
(1.School of Tropical Agriculture and Forestry, Hainan University, Haikou 570228, China;2.School of Computer Science and Technology, Hainan University, Haikou 570228, China)
|
| Abstract: |
| Metabolite-Protein Interaction (MPI) is crucial in regulating protein functions and controlling various cellular processes. In view of the problem that the traditional method of detecting MPI using experimental methods and high-throughput metabolome-transcriptome joint analysis leads to high material and time costs, it is proposed to apply the emerging heterogeneous graph attention network in the field of deep learning to plant MPI prediction. Therefore, in order to predict the MPI of rice, it is proposed to establish a rice metabolite-protein association prediction model OryHNCGAT (Oryza Heterogeneous Neighbor Contrastive Graph Attention network). The OryHNCGAT model aggregates information at the neighbor level and the relationship level by introducing an attention mechanism to obtain a more comprehensive node representation. Secondly, heterogeneous neighbor contrastive learning is added, and the learned node embedding can well preserve the heterogeneous network topology. The experimental results show that the OryHNCGAT model achieves effective rice metabolite-protein association prediction, providing a new method and approach for further studying metabolite-protein interactions. |
| Key words: Neural network Heterogeneous graph attention network Heterogeneous neighbor contrastive learning Oryza Metabolite-protein interaction |
