CASIA OpenIR  > 毕业生  > 博士学位论文
基于复杂网络的学习模型及在中医药领域的应用研究
Alternative TitleResearches on Complex Network-based Learning Models and Their Applications in Traditional Chinese Medicine
宋江龙
Subtype工学博士
Thesis Advisor杨一平
2015-05-29
Degree Grantor中国科学院大学
Place of Conferral中国科学院自动化研究所
Degree Discipline计算机应用技术
Keyword复杂网络 中医药 模块检测 症状网络 关系推理 Complex Network Traditional Chinese Medicine Module Detection Symptom Network Relation Reasoning
Abstract本文的研究主要是利用复杂网络来模拟中医药领域的真实系统,并提出基于网络模型的分析预测方法,解决中医药领域的相关问题。从“方剂”的角度,本文提出基于分子网络的预测方法,用于预测中药复方的有效成分并且挖掘其作用机制;从“证候”的角度,本文采用症状网络来研究同病异证的临床规律,而且搭建了基于多层关联网络的中医诊疗辅助决策的原型系统。本文的主要研究内容包括以下几个方面: (1)基于分子网络的复方作用机制挖掘 针对高通量试验的研究成本高、周期较长的问题,本文提出基于分子网络的复方作用机制挖掘方法。首先,根据一个复方的中药组成,利用现有的生物医学资源,收集复方的化学成分和潜在基因靶标,构建复方的基因交互网络。其次,采用GN算法识别基因网络中的显著模块,并筛选核心模块进行富集分析。然后,根据中药成分关联的潜在靶标,计算成分-通路得分;根据复方治疗的疾病致病基因,并计算疾病-通路得分。最后,结合成分-通路与疾病-通路得分,计算中药成分与疾病的关联得分,以此预测复方的有效成分,并通过有效成分作用的富集通路分析复方的作用机制。本文先收集帕金森氏病的治疗药物以及其它无关药物,构建药物对应的靶标网络,采用该方法预测帕金森氏病的有效药物。预测结果表明该方法是有效的。随后,该方法被应用于治疗流感的经典方——疏风解毒胶囊,利用分子网络挖掘其作用机制。现有的生物医学文献证实了大部分的预测和分析结果,间接证明该方法的有效性。 (2)基于混合模块度的模块检测算法 根据“相似的化合物有相似的性质”这一思想,本文为复方的分子系统引入成分相似关系。于是,复方的分子系统就能够采用包含成分相似关系、成分-靶标交互和基因交互关系的二类异质网络来模拟。针对现有的算法在检测二类异质网络的模块时存在的不足,本文提出了基于混合模块度的模块检测算法MixMod。MixMod是一种采用快速优化策略的层次聚类算法,主要包括局部优化和模块聚合两个步骤。MixMod算法以混合模块度为优化目标,能够快速检测二类异质网络的模块结构。通过在人工基准网络和真实基准网络中的测试,我们发现MixMod算法的模块检测性能比CNM、MCL、Infomap和Louvain这四种经典算法更优越。最后,MixMod被应用于步长脑心通胶囊(BNC)的成分-基因异质网络,挖掘其分子作用机制。而采用BNC肠吸收液的体外试验证实了之前预测结果的准确性。这说明MixMod是检测二类异质网络模块结构的一种有效算法,能够用于挖掘中药复方的作用机制。 (3)基于症状网络的同病异证研究 针对中医临床的“同病异证”现象,本文提出了基于症状网络的同病异证规律分析方法。首先,对依据中医量表采集的临床数据进行预处理,同时分析疾病的主要症状特点。其次,采用Fisher得分衡量症状对疾病下某种证候的分类能力,并结合SVM辨证模型筛选证候的核心症状。然后,利用证候的“阳性”样本集和核心症状,构建此证候的症状网络。最后,通过置换检验识别每种证候的网络中存在的动态症状关联。应用该方法分析抑郁症的临床数据,研究抑郁症和它的7种证候在临床症状组成、症状网络和...
Other AbstractThe work presented in this paper is essentially to model real systems in Traditional Chinese Medicine (TCM) using complex networks and to propose different network-based approaches, in order to solve practical problems in TCM research. In the research of TCM formula, we proposed an approach employing molecular networks to predict the effective components of a TCM formula and then uncover its mode of action. In the research of syndrome, we used symptom network to study the clinical rules of different syndromes in same disease, and built a decision support system for TCM diagnosis and treatment based on multilevel association networks. More specifically, the work of this paper involves several aspects as introduced below. (1). Uncovering the mode of action of a TCM formula. Due to the high cost and long period of high-throughput assays, we proposed an approach to uncover the mode of action of a TCM formula based on its molecular network. First, according to the herbal composition of a TCM formula, all chemical components in each herb were collected and the potential targets associated with these chemicals were also retrieved from pharmacological databases. The interactions between potential targets were subsequently extracted from PPI (protein-protein interaction) databases. A gene network for the formula could be constructed. Second, Girvan-Newman algorithm was employed to detect modules from the gene network, and key modules were studied through enrichment analysis. Third, we computed the association scores between chemicals and enriched pathways via common genes and interactions. The association scores between diseases and pathways were computed in a similar way. Fourth, we predicted effective components of the formula by combining chemical-pathway and disease-pathway associations, and investigated the molecular mechanism through enriched pathways related to effective components. Finally, this approach was used to predict effective drugs for Parkinson's disease, and then uncover the mode of action of Shufeng jiedu capsule (an anti-influenza formula). The results of both experiments demonstrated that this approach is accurate and effective to uncover the molecular mechanism of TCM formulae. (2). A module detection algorithm based on mixed modularity. Based on the idea of “chemicals with similar structures have similar properties”, we introduced chemical similarity to the molecular system of a TCM formula. Thus, the molecular system could be model...
Other Identifier201218014629096
Language中文
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/6728
Collection毕业生_博士学位论文
Recommended Citation
GB/T 7714
宋江龙. 基于复杂网络的学习模型及在中医药领域的应用研究[D]. 中国科学院自动化研究所. 中国科学院大学,2015.
Files in This Item:
File Name/Size DocType Version Access License
CASIA_20121801462909(2512KB) 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[宋江龙]'s Articles
Baidu academic
Similar articles in Baidu academic
[宋江龙]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[宋江龙]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.