基于SVM的Aβ聚集抑制剂和血脑屏障通透性预测模型

发布时间：2018-09-12 12:34

【摘要】：随着社会的不断发展,物质和经济条件的提高,人口老年化程度越来越高,阿尔兹海默症(Alzheimer's disease, AD)患者的数量在全球也迅速的增加,给全球的经济、社会以及部分家庭带来了沉重的负担,因此研发能够治疗AD的药物变得非常迫切。由于p淀粉样蛋白(β-amyloid, Aβ)代谢过程在AD的发生中具有很重要的作用,如果能有效地控制Aβ的聚集将有望从根本上治疗AD。本文首先构建了能够筛选Ap聚集抑制剂的支持向量机(Support Vector Machine, SVM)模型,然后用上述模型从中药化学数据库(Traditional Chinese Medicines Database, TCMD)中筛选出合适的中药,最后用作者自己构建的血脑屏障(Blood Brain Barrier, BBB)通透性模型测试上述中药的通透性,最终确立具有Aβ抑制效益并能通过血脑屏障的中药,进而为以后的实验验证以及进一步的药物研发做好基础和提供一些参考价值。本文中总共构建了四个SVM模型,包括两个Aβ40小分子抑制剂模型、一个Aβ42小分子抑制剂模型和一个BBB模型,由于Aβ40聚集抑制剂的数据涉及到两种不同的实验self seeding radio assay与self seeding thioflavin T assay,为了区别这两种数据建立的模型,便分别称为Aβ40聚集抑制剂预测模型R和Aβ40聚集抑制剂预测模型T。构建BBB通透性预测模型所用的样本数量为415,训练集的准确度为92.5%,测试集的准确度为79.4%,五重交叉验证得到的平均准确率为80.3%。构建Aβ40聚集抑制剂预测模型R所使用的样本数量为64,训练集的准确度为88.6%,测试集的准确度为75.0%,留一法验证的平均准确率为75.0%。从TCMD23033个化合物中总共筛选出了41个化合物,然后统计出含有这些化合物的种类大于等于3种的中药,结果为胡椒属(Piper chaba)(果实)、雷公藤(Tripterygium wilfordii)干姜(Zingiber officinale)、高良姜(Alpinia officinarum)和益智仁(Alpinia oxyphylla)。最后这些中药经过BBB模型筛选后,胡椒属、高良姜和益智仁符合模型的要求。建立Aβ40聚集抑制剂预测模型T使用的样本数量为82,训练集的准确度为75.0%,测试集的准确度为57.7%,留一法验证的平均准确率为71.4%,从TCMD中总共筛选出了100个化合物,然后统计出含有这些化合物的种类大于等于3种的中药,结果为胡椒属西洋参(Panax quinquefolium)、麝香(Moschus moschiferus)、壮观番泻(Senna spectabilis)(花)、高良姜(Alpinia officinarum)、蒜臭母鸡草(Petiveria alliacea)(根,茎和叶)、益智仁(Alpinia oxyphylla)和金鸡勒(Cinchona ledgeriana)。最后这些中药经过BBB模型筛选后,所有的中药均符合模型的要求。建立Aβ42聚集抑制剂预测模型所需样本数量为28,训练集的准确度为95.0%,测试集的准确度为62.5%,留一法验证的平均准确率为55.0%,从TCMD2303个化合物中仅筛选出了1个化合物,为芸苔子(Brassica campestris)的成分。最后经过BBB模型筛选后,其符合模型的要求。
[Abstract]:With the continuous development of society and the improvement of material and economic conditions, the degree of population aging is becoming higher and higher. The number of (Alzheimer's disease, AD) patients with Alzheimer's disease is also increasing rapidly in the world, giving the global economy. Society and some families have a heavy burden, so it is urgent to develop drugs to treat AD. As the metabolic process of 尾 -amyloid (A 尾) plays an important role in the development of AD, it is expected that the aggregation of A 尾 can be effectively controlled in the treatment of AD.. In this paper, the support vector machine (Support Vector Machine, SVM) model was constructed to screen Ap aggregation inhibitors, and then the suitable Chinese medicine was selected from the traditional Chinese medicine database (Traditional Chinese Medicines Database, TCMD) by using the above model. Finally, the (Blood Brain Barrier, BBB) permeability model of blood-brain barrier was used to test the permeability of the above-mentioned Chinese medicine, and the Chinese medicine with A 尾 inhibitory effect was finally established, which could pass through the blood-brain barrier. Further more, it can provide some reference value for the future experimental verification and further drug research and development. In this paper, four SVM models are constructed, including two A 尾 40 small molecular inhibitor models, one A 尾 42 small molecule inhibitor model and one BBB model. Because the data of A 尾 40 aggregation inhibitor involve two different experimental models, self seeding radio assay and self seeding thioflavin T assay, in order to distinguish the two models, they are called A 尾 40 aggregation inhibitor prediction model R and A 尾 40 aggregation inhibitor prediction model T respectively. The number of samples used to construct the BBB permeability prediction model is 415, the accuracy of the training set is 92.5, the accuracy of the test set is 79.4, and the average accuracy of the five-fold cross-validation is 80.3. The number of samples used to construct the A 尾 40 aggregation inhibitor prediction model R was 64, the accuracy of the training set was 88.6, the accuracy of the test set was 75.0, and the average accuracy of the remaining method was 75.0. A total of 41 compounds were screened out of the TCMD23033 compounds, and the Chinese medicines containing more than 3 of them were identified, the result being (Piper chaba) (fruit.), (Tripterygium wilfordii), (Tripterygium wilfordii), ginger, (Zingiber officinale), ginger, (Alpinia officinarum), and Ishihiren, (Alpinia oxyphylla). Finally, after the BBB model was selected, pepper, alpinia and Izhiren met the requirements of the model. The number of samples used to establish the A 尾 40 aggregation inhibitor prediction model T was 82, the accuracy of the training set was 75.0, the accuracy of the test set was 57.7, the average accuracy of the remaining method was 71.4. A total of 100 compounds were screened from the TCMD. Then the Chinese medicines containing more than 3 kinds of these compounds were found. The results were: (Panax quinquefolium), musk (Moschus moschiferus), of Panax quinquefolium, spectacular (Senna spectabilis) (flower), (Petiveria alliacea) (root, stem and leaf of (Alpinia officinarum), garlic hens), (Alpinia oxyphylla) and (Cinchona ledgeriana). Finally, after the BBB model was selected, all the traditional Chinese medicines met the requirements of the model. The number of samples needed to establish the A 尾 42 aggregation inhibitor prediction model was 28, the accuracy of the training set was 95.0, the accuracy of the test set was 62.5, the average accuracy of the remaining method was 55.0. Only one compound was screened out of the TCMD2303 compounds. It is a component of (Brassica campestris) from Brassica brassica. Finally, after BBB model screening, it meets the requirements of the model.
【学位授予单位】：华中农业大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R-332

【相似文献】