摘要
目的阐明物质成分与中药药性间的内在联系和定量关系,建立物质成分与药性之间的统计模式识别模型,确定药性特征标记物质,实现利用物质成分识别和预测中药药性。方法基于高效液相色谱数据,在数据预处理基础上,利用偏最小二乘判别(PLS-DA)模型建立色谱数据的中药药性统计模式识别模型,并比较不同预处理的变量标准化或转换对模型精度的影响。结果变量经过均值标准化的Ctrl/PLS-DA模型不仅具有很高的判别率(100%)和较高的预测准确度(92.31%),且由其参数确定的药性特征标记适宜解释中药药性的寒热特征。结论色谱数据的预处理可以有效降低数据维度,抑制噪声,提高数据质量。运用PLS-DA建立的中药药性识别模型具有较高的判别正确率和预测精度,根据模型参数可以确定表征中药寒热药性的特征标记,为进一步的临床或动物实验验证提供理论依据。
Objective To clarify internal relation and quality relationship between the material composition and the Chinese herbal medicine property (CHMP).To establish statistic pattern recognition model for identifying the Chinese herbal medicine property markers (CHMP-markers),and achieve to classify and forecast the CHMP by means of the material composition.Methods Based on preprocessing technique,Partial Least Squares Discriminant Analysis (PLS-DA) was selected to build the CHMP recognition model from High Performance Liquid Chromatography (HPLC) data.The influence on model precision by different variable transformation was analyzed.Results The Ctrl/PLS-DA model using the variable mean normalization method had high discriminate ratio (100%) and prediction accuracy (92.31%).The CHMP-markers determined by this model can appropriately explain the features of CHMP.Conclusion Processing of spectra data can effectively reduce data dimension,control the noise and improve the data quality.The CHMP recognition model building by PLS-DA has high classification and prediction accuracy.The CHMP-markers can be determined by model's parameters,in addition to provide a guidance for the experimental verification next step.
出处
《中国卫生统计》
CSCD
北大核心
2011年第6期628-631,637,共5页
Chinese Journal of Health Statistics
基金
国家"973"项目:中药药性理论相关基础问题研究(2007CB512601)
关键词
中药药性标记
高效液相色谱
数据预处理
偏最小二乘判别分析
Chinese herbal medicine property markers
High performance liquid chromatography
Data preprocessing
Partial least squares discriminant analysis