【作者】 王远强；

【导师】 李志良；

【作者基本信息】 重庆大学 ， 药物化学， 2005， 硕士

【摘要】 中医药(TCM)的发展历史与人类文明史同步,是中国民族智慧的结晶,为人类的健康做出了巨大贡献。中医药在几千年的发展中,已形成了完善的理论体系、积累了丰富的医疗实践经验,对现代医药发展具有积极的推动作用。近年来,中医药受到众多国家与地区的重视与肯定,并被积极开发利用。目前,众多中药创新药物已被研制开发。可以预计,以天然中药为基础开发创新药物将成为21世纪医药发展的重要领域。定量结构活性相关(QSAR)是现代化学基础研究与应用的重要领域,已成为有机化学、药物化学、环境化学、计算化学、农药乃至分子生物学、免疫学的研究热点,但在中药复方制剂研究以及中药活性成分筛选中的应用却比较有限。本文使用分子电性距离矢量(MEDV)表征中药活性成分分子结构,并用多元线性回归(MLR)、逐步回归(STR)、主成分分析(PCA)、偏最小二乘回归(PLS)与人工神经网络(ANN)等方法研究中药活性成分的结构参数与其药效之间的相关关系,建立QSAR模型并用于未知药效的活性成分的预测。本文研究分为以下几个部分: 1. 使用STR与PLS对抗SARS中药活性成分进行药效识别研究,用STR筛选显著结构参数建立各种药效的相关模型,对测试集样本的预测率均在80%以上;用PLS方法对退热、抗炎、改善血循环、预防等药效建立的相关模型对测试集样本的预测率也均在80%以上,两种方法所建立的模型具有较好的预测能力。2. 选取对人体鼻咽癌(KB)细胞有致毒活性的中药活性成分进行QSAR研究。分别用STR和PLS方法建立结构参数与药效的QSAR模型,其最优模型的复相关系数(R2)分别为0.674与0.618;交互校验复相关系数(Q2)分别为0.519与0.547;对测试集样本的均方根误差(RMS)分别为1.488与1.434,所建立的模型具有良好的预测能力。3. 用人工神经网络(ANN)以及主成分神经网络(PCA-ANN)对抗SARS中药活性成分进行药效识别研究。两种方法建立的模型对测试集样本的退热、抗炎、改善血循环、预防等药效的预测率均高于85%。研究结果表明:PCA-ANN模型对各种药效的预测能力与ANN模型相当,但对退热药效的预测能力有明显改进。4. 使用ANN与PCA-ANN对有抗肿瘤活性的活性成分进行QSAR研究,两种方法所建模型对测试集样本的RMS分别为1.530与1.516。研究结果表明:ANN模型与PCA-ANN模型的预测能力相当,但明显不如STR与PLS模型。5. 使用ANN与PCA-ANN对有致毒活性的中药活性成分进行识别研究,两种方法所建模型的预测率分别为64.00%与73.00%。研究结果表明:两种方法所建立的模型对中药活性成分的筛选具有一定指导作用,但还需要进一步改进。更多还原

【Abstract】 Traditional Chinese Medicine （TCM）, whose development history is in accordance with human history of civilization, embodies intelligence of the Chinese people and has made great contribution to human health. In the long course of its development, TCM has developed its own theoretical system and accumulated plenty of experiences, which has greatly promoted the development of modern medicine. Recently, more and more countries have realized its value, and numerous innovative drugs have been developed with vigorous efforts. Based on abundant natural resources of TCM, the exploitation of innovative drugs and screening of active components will become an important field in the development of medicine in 21th century. Quantitative structure activity relationship （QSAR） is an important field for basic research and application of modern chemistry. Although it has become the hotspot of research on organic chemistry, pharmacological chemistry, environment chemistry, chemometrics, molecular biology and immunology, it’s inadequate in the research on active components screening of TCM. In this paper, Molecular electro-negativity distance vectors （MEDV） is adopted to express the structure of active components of TCM. Such methods as MLR, STR, PCA, PLS and ANN were used to find the relationship between structure parameters and activity of active components in certain TCM. Models were established with good performance in predicting unknown active components. The research can be divided into the following sections: 1. First of all, MEDV was employed to characterize molecular structure of active components in TCM against SARS and STR to select the important parameters. STR was used to set up correlative models of structure parameters and various activities. The predictive capability using these models for the testing samples were over 80%. While the predictive capability of PLS were over 80% for defervesce, anti-inflammation, promoting blood circulation and prophylactic effect. 2. TCM with toxicity to KB cell was selected to study with QSAR. Models were established by STR and PLS, constructing the relationship between structural parameters and -log(ED₅₀). The correlative coefficients （R²） of the optimized model were 0.674 and 0.618; the correlative coefficients of cross validation （Q²） were 0.519 and 0.547; RMS of testing samples were 1.488 and 1.434. The result showed that the model owned good predictive capability. 3. ANN and PCA-ANN was adopted to distinguish drug effect of anti-SARS active components in TCM. Predictive capabilities for the testing samples were over 80% and 85% of ANN and PCA-ANN for defervesce, anti-inflammation, promoting blood circulation and prophylactic effect. 4. ANN and PCA-ANN were adopted to investigate active components with toxicity to KB cell. RMS of the two methods were 1.530 and 1.516 respectively. The results indicated that PCA-ANN is a better method for prediction than ANN. 5. ANN and PCA-ANN were adopted to distinguish active components with toxicity for KB cell. The predictive capabilities were 64.00% and 73.00% respectively. To some extent, the results illustrated both methods can contribute to the screening of active components. 6. ANN and PCA-ANN were adopted to distinguish the active components which can cause acute toxicity. The results show that PCA-ANN gets a better result than ANN. Establishing models through QSAR method can help to screen active components in TCM and are a new way for TCM research. QSAR research for TCM will have a bright further.更多还原