基于中药资源的计算机辅助药物分子设计

发布时间：2018-07-25 12:50

【摘要】：近年来，随着越来越多的天然产物成功地通过FDA认证而上市，中药（Traditional Chinese Medicines，TCMs）作为天然产物的重要组成部分，在现代药物研发中受到了越来越多的关注和重视。中药用于治疗疾病的主要形式是通过含有多种中草药植物的中药复方来实现的，因此人们普遍认为，中草药可以作为药物研发很好的类药化合物来源。从传统中草药中寻找到相关靶点的潜在活性化合物并确定其药理活性已经成为制药公司药物开发的一个重要途径。人们对基于中草药资源的药物研发已经做过了大量的尝试和研究，但我们对中草药化合物的分子的性质、结构以及成药性特征还缺乏深入的了解。此外，相比较于西药治病理论，大部分中草药治疗疾病的机制都还不够清晰，能否从分子水平阐述中草药治疗相关疾病的作用机制是非常重要的研究课题。最后，如何从中草药化合物中筛选得到相关靶点的潜在活性化合物也是一个热点研究方向。本论文系统开展了基于中草药有效成分的计算机辅助药物分子设计研究。首先，我们系统比较了药物数据库MDDR、非药数据库ACD和中草药化合物数据库（TCMCD）中化合物的物理化学性质以及结构特征的差异。结果表明，相比MDDR和ACD，TCMCD中的化合物性质分布更为广泛并且结构更为复杂和新颖。同时，我们发现基于简单性质的类药性预测规则预测能力较差。为了对中草药化合物的类药性进行定量评价，我们用机器学习方法，包括朴素贝叶斯和递归分割方法，构建了精确的类药性定量预测模型。结果表明，基于分子理化性质描述符构建的类药性模型的预测精度较低，而引入了分子指纹描述符后，类药性模型的预测精度有了较大的提升。同时，我们发现类药性模型的预测能力与训练集的大小以及构成有着直接的关系，用所构建的最为可靠的类药性模型对中草药化合物数据库进行了类药性的评价，超过60%的中草药化合物被预测为类药，表明TCMCD从统计上讲是类药的，可以作为药物研发的一个很好的类药化合物来源。中药治疗疾病主要是通过由多种中草药植物所组成的中药复方的形式发挥作用，因此，由大量中药有效成分构成的中药复方的治疗疾病的机制很不清晰。为了从分子水平阐述中草药复方治疗疾病的机制，我们以治疗Ⅱ型糖尿病中药复方为例进行研究。首先，收集已知治疗Ⅱ型糖尿病的中药复方中含有的有效成分化合物以及与Ⅱ型糖尿病相关的靶点。随后采用分子对接、药效团映射以及机器学习的方法筛选出各靶点的潜在活性化合物。通过构建潜在活性化合物和靶点的相互作用网络，从一定程度上揭示了中草药复方治疗Ⅱ型糖尿病的机制：中药复方中的大部分有效成分只能跟单一靶点形成相互作用，构成治疗Ⅱ型糖尿病的主要作用力，其次，中药复方中的少部分化合物能和多个Ⅱ型糖尿病相关靶点作用，发挥治疗糖尿病的次要作用，协同增强治疗糖尿病的效果，最后，中草药中的部分化合物不与Ⅱ型糖尿病相关靶点形成直接的相互作用，而是通过其他的一些药理活性，，如去自由基功能/抗氧化能力、抗菌能力来协助治疗糖尿病及其并发症。所得到的这些结论能够较好的与经典中医药治病理论“君臣佐使”相吻合。为了从中草药化合物数据库TCMCD中筛选得到相关靶点理想的潜在活性化合物，我们以激酶靶点ROCK1为例展开研究。考虑到蛋白柔性对虚拟筛选结果的影响，我们用机器学习方法整合ROCK1靶点多个复合物结构所得到的分子对接和药效团模型的预测结果，构建了新颖的并行虚拟筛选策略并对其预测能力进行了评测。研究结果表明，相比较于基于单个复合物结构的分子对接或药效团模型的预测结果，整合的虚拟筛选策略更为可靠。随后，用构建的并行虚拟筛选策略对中草药化合物数据库进行了虚拟筛选，得到了53个结构新颖的ROCK1潜在活性化合物。这些化合物可以作为理想的ROCK1潜在活性化合物来进行后续的研究。所构建的并行虚拟筛选策略也可以作为一个可靠的工具用于药物筛选。
[Abstract]:In recent years, as more and more natural products have been successfully listed by FDA certification, Traditional Chinese Medicines (TCMs), as an important component of natural products, has attracted more and more attention and attention in modern drug research and development. The main form of Chinese medicine for the treatment of diseases is through a variety of herbal plants. It is widely believed that Chinese herbal medicine can be used as a good source of drugs for drug development. It is an important way for pharmaceutical companies to develop potential active compounds from traditional Chinese herbal medicine and determine their pharmacological activities. People are on the basis of Chinese herbal medicine resources. There has been a lot of research and Research on drug research and development, but we do not know much about the properties, structure and characteristics of the molecules of Chinese herbal compounds. In addition, compared with the western medicine treatment theory, the mechanism of most Chinese herbal medicines for the treatment of diseases is not clear enough to explain the correlation of Chinese herbal medicine at the molecular level. The mechanism of the action of the disease is a very important research topic. Finally, how to screen the potential active compounds from Chinese herbal medicine compounds is also a hot research direction.
In this paper, a computer aided drug molecular design based on the effective components of Chinese herbal medicine is systematically carried out. First, we systematically compare the physical and chemical properties and structural characteristics of the compounds in the drug database MDDR, the non drug database ACD and the Chinese herbal compound database (TCMCD). The results show that compared to MDDR and ACD, TCMCD The properties of the compounds are more widely distributed and more complex and novel. At the same time, we have found that the prediction rule of the drug resistance prediction rules based on simple properties is poor. In order to evaluate the drug resistance of Chinese herbal compounds, we use machine learning methods, including the simple Juliu and the recursive segmentation method, to construct an accurate class. The results showed that the prediction accuracy of the model based on molecular physicochemical descriptors was low, and the prediction accuracy of the model was greatly improved after introducing the molecular fingerprint descriptor. At the same time, we found that the pretest ability of the drug class model and the size of the training set and the composition were straight. The relationship was evaluated with the most reliable model of drug resistance in the Chinese herbal compound database. More than 60% of the Chinese herbal compounds were predicted to be drug classes, indicating that TCMCD is a statistical class of drugs and can be used as a good source of drug class compounds in drug development.
Chinese medicine for the treatment of diseases is mainly through the form of Chinese herbal compound made up of a variety of Chinese herbal medicines. Therefore, the mechanism of the treatment of disease by a large number of effective ingredients of Chinese medicine is not clear. First, we collect effective compounds and targets related to type II diabetes, and then use molecular docking, pharmacophore mapping and machine learning to screen out potential active compounds from each target. By constructing potential active compounds and targets. The point interaction network reveals the mechanism of Chinese herbal compound treatment for type II diabetes to a certain extent: most of the effective components in the Chinese herbal compound can only interact with a single target and constitute the main force for the treatment of type II diabetes. Secondly, a few compounds in the Chinese herbal compound can be associated with multiple type of type 2 diabetes. Closing the target point to play a secondary role in the treatment of diabetes and synergistically enhance the effect of diabetes. Finally, some of the compounds in Chinese herbal medicine do not interact directly with the related targets of type 2 diabetes, but by other pharmacological activities, such as free radical function / antioxidant capacity, and antibacterial ability to assist in the treatment of diabetes. Disease and its complications. These conclusions can be better consistent with the classic Chinese medicine theory.
In order to screen the potential active compounds of the target target from the Chinese herbal compound database TCMCD, we take the kinase target ROCK1 as an example. Considering the effect of the protein flexibility on the virtual screening results, we use the machine learning method to integrate the molecular docking and the pharmacophore of multiple complex structures of the target of ROCK1. A novel parallel virtual screening strategy is constructed and its prediction ability is evaluated. The results show that the integrated virtual screening strategy is more reliable compared to the prediction results of molecular docking based on single complex structure or the model of the pharmacophore. Subsequently, the constructed parallel virtual screening strategy is used in the middle of the model. The herbal compound database has been virtual screening, and 53 novel ROCK1 potential active compounds are obtained. These compounds can be used as ideal ROCK1 potential active compounds for subsequent research. The parallel virtual screening strategy can also be used as a reliable tool for drug screening.
【学位授予单位】：苏州大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：R91-39

【相似文献】