中国组织工程研究 ›› 2025, Vol. 29 ›› Issue (35): 7552-7561.doi: 10.12307/2025.968

• 组织构建实验造模 experimental modeling in tissue construction • 上一篇    下一篇

基于人工智能和组学数据驱动的中药潜在机制新型分析预测方法

江启煜1,曾慧妍2   

  1. 1广州中医药大学,广东省广州市  510006;2广东省中医院,广东省广州市  510120

  • 收稿日期:2024-11-15 接受日期:2024-12-23 出版日期:2025-12-18 发布日期:2025-05-06
  • 通讯作者: 曾慧妍,博士,主任医师,广东省中医院,广东省广州市 510120
  • 作者简介:江启煜,男,1984年生,广东省广州市人,汉族,硕士,讲师,主要从事中医药人工智能的研究。
  • 基金资助:
    国家自然科学基金项目(82374233),项目负责人:曾慧妍;广东省自然科学基金(2414050003181),项目负责人:曾慧妍

A novel analysis and prediction method for potential mechanisms of traditional Chinese medicine based on artificial intelligence and omics data-driven approach

Jiang Qiyu1, Zeng Huiyan2   

  1. 1Guangzhou University of Chinese Medicine, Guangzhou 510006, Guangdong Province, China; 2Guangdong Provincial Hospital of Traditional Chinese Medicine, Guangzhou 510120, Guangdong Province, China
  • Received:2024-11-15 Accepted:2024-12-23 Online:2025-12-18 Published:2025-05-06
  • Contact: Zeng Huiyan, MD, Chief physician, Guangdong Provincial Hospital of Traditional Chinese Medicine, Guangzhou 510120, Guangdong Province, China
  • About author:Jiang Qiyu, Master, Lecturer, Guangzhou University of Chinese Medicine, Guangzhou 510006, Guangdong Province, China
  • Supported by:
    National Natural Science Foundation of China, No. 82374233 (to ZHY); Natural Science Foundation of Guangdong Province, No. 2414050003181 (to ZHY)

摘要:


文题释义:
空间转录组(Spatial Transcriptomics):是一种结合了组织切片技术、单细胞测序技术和传统转录组学的高新技术。它能够在保持组织空间结构的同时,分析基因在特定空间位置的表达情况,从而提供高分辨率的基因表达信息。空间转录组技术旨在分析基因在组织特定空间位置的表达量。
深度自编码神经网络(Deep Autoencoders):是一种特殊的神经网络架构,属于无监督学习的一种,其核心目标是通过对输入数据进行压缩和重建,学习输入数据的本质特征。深度自编码器通常由多层编码器和解码器组成。

背景:中药对疾病的治疗是一个复杂的多靶点调控过程。如何利用人工智能、单细胞转录组、空间转录组以及生物信息学等多个领域的技术相结合,探索中药的多靶点整合效应具有重要意义。
目的:基于人工智能和组学数据驱动提出一种有别于网络药理学的中药潜在机制新型分析预测方法,并以探索大柴胡汤治疗高脂血症及动脉粥样硬化的潜在机制为例。
方法:①通过TCMSP数据库收集大柴胡汤组成药物的药效蛋白靶点,在Genecards、NCBI、TTD等数据库获取高脂血症的疾病靶点。②从GEO数据库获取高脂血症单细胞转录组[第一组为野生型(WT)、Apoe基因敲除、Ldlr基因敲除小鼠的主动脉瓣单细胞数据样本;第二组为Ldlr基因敲除小鼠高胆固醇喂食与正常喂食的单细胞数据样本]及人冠脉粥样硬化组织切片空间转录组样本。构建深度计数自编码网络,将转录组测序数据进行编码,并利用单细胞转录组及空间转录组技术将整合编码值(MTIS)映射到单细胞水平及空间组织水平上,进行样本对比统计分析,并进行主要效应细胞与效应基因的识别。
结果与结论:①大柴胡汤对WT型与Apoe基因敲除型小鼠之间、WT型与Ldlr基因敲除型小鼠之间的MTIS均存在数据形态以及统计学上的差异(P < 0.000 1);②大柴胡汤对Apoe基因敲除型小鼠的潜在效应细胞是主动脉瓣间质细胞,而对Ldlr基因敲除型小鼠的潜在效应细胞是白细胞、纤维细胞、血管内皮细胞;潜在效应基因是Vcam1、Fn1、Mmp2;③大柴胡汤对Ldlr敲低型小鼠高胆固醇喂食样本与正常喂食样本的MTIS存在数据形态以及统计学上的差异(P < 0.000 1),潜在效应细胞是巨噬细胞;潜在效应基因是Fn1、F7、Ptgs1、IL6、App;④人类冠脉切片的空间转录组MTIS对比表明,MTIS高值细胞似乎在血管以及硬化斑块区域都有分布,而MTIS低值细胞似乎主要集中于血管内皮以及硬化斑块区域等病变区域。结论:该新型分析方法实现了单细胞水平以及器官空间组织水平上中药多靶点整合潜在效应的量化分析,探索了大柴胡汤治疗高脂血症及动脉粥样硬化的潜在机制。
https://orcid.org/0009-0002-4358-0212(江启煜)

中国组织工程研究杂志出版内容重点:组织构建;骨细胞;软骨细胞;细胞培养;成纤维细胞;血管内皮细胞;骨质疏松;组织工程

关键词: 人工智能, 单细胞转录组, 空间转录组, 生物信息学, 网络药理学, 多靶点, 中药, 基因敲除, 组学, 药理

Abstract: BACKGROUND: The treatment of diseases with traditional Chinese medicine is a complex multi-target regulatory process. It is of great significance to explore the multi-target integration effect of traditional Chinese medicine by combining technologies from multiple fields such as artificial intelligence, single-cell transcriptomics, spatial transcriptomics, and bioinformatics. 
OBJECTIVE: To propose a novel analytical prediction method for potential mechanisms of traditional Chinese medicines, which is different from network pharmacology, based on artificial intelligence and omics data driven, with an example of exploring the potential mechanisms of Dachaihu Decoction for treating hyperlipidemia and atherosclerosis.
METHODS: (1) The pharmacodynamic protein targets of the constituent drugs of Dachaihu Decoction were collected through TCMSP database, and the disease targets of hyperlipidemia were obtained in Genecards, NCBI, and TTD. (2) The single-cell transcriptome samples of hyperlipidemia (the first set of single-cell data samples from aortic valves of wild-type, Apoe knockout, and Ldlr knockout mice; the second set of single-cell data samples from Ldlr knockout mice fed with high cholesterol versus normal feeding) and spatial transcriptome samples from human coronary atherosclerosis tissue sections were obtained from the GEO database. A deep neural network autoencoder model was developed to encode the transcriptome sequencing data, and the integrated coded values (MTIS) were mapped to the single-cell level and spatial organization level using single-cell transcriptome and spatial transcriptome technologies for comparative statistical analyses of the samples and identification of the main effector cells and effector genes. 
RESULTS AND CONCLUSION: (1) There were significant differences in the data morphology and statistics of MTIS between wildtype and Apoe-knockout mice treated with Dachaihu Decoction (P < 0.000 1), as well as between wildtype and Ldlr-knockout mice treated with Dachaihu Decoction (P < 0.000 1). (2) The main effector cells of Dachaihu Decoction in Apoe-knockout mice were aortic valve stromal cells, while the main effector cells in Ldlr-knockout mice were white blood cells, fibroblasts, and vascular endothelial cells. Except for Ldlr and Apoe, the main effector genes are Vcam1, Fn1, and Mmp2. (3) There were statistically significant differences (P < 0.000 1) in MTIS between high cholesterol fed samples and normal fed samples of Ldlr-knockout mice treated with Dachaihu Decoction. The main effector cells were macrophages, and the main effector genes were Fn1, F7, Ptgs1, IL6 and App. (4) The spatial transcriptome comparisons of MTIS in human coronary artery slices showed that high MTIS value cells appeared to be distributed in both blood vessels and atherosclerotic plaque areas, while low MTIS value cells appeared to be mainly concentrated in the endothelial cells and atherosclerotic plaque areas. To conclude, this new analytical method achieves quantitative analysis of the multi-target integration effects of traditional Chinese medicine at the single-cell level and organ spatial tissue level, which is used to explore the potential mechanism of Dachaihu Decoction in treating hyperlipidemia and atherosclerosis.

Key words: artificial intelligence, single cell transcriptome, spatial transcriptome, bioinformatics, network pharmacology, multi-target, traditional Chinese medicine, knockout, omics, pharmacology

中图分类号: