基于60种植物药多糖HPLC图谱的寒热药性分析及统计模式识别研究

Study on appropriate statistical pattern recognition model for relationship between cold-hot natures and polysaccharides based on HPLC of 60 traditional Chinese medicines

目的:探讨中药寒热药性与多糖成分的相关性,寻找适合解释此相关性的统计模式识别方法。方法:选取寒、热性植物药各30种,提取和精制多糖并彻底水解成单糖,进行衍生化反应,测定多糖的单糖组成HPLC指纹图谱并构建数据库,经数据预处理后,比较6种统计模式识别方法在识别中药药性特征标记方面的优劣。结果:与其他模型相比,偏最小二乘判别分析识别率最高,对测试集植物药的组内判别正确率为91.7%,且对96个模拟药性特征标记整体识别率也明显优于其他模型。结论:中药寒热药性与多糖成分有明显相关性,用偏最小二乘判别分析建立的中药药性识别模型最适合发现与解释中药多糖成分与药性之间的相关性。

Objective： To discuss the correlation between cold-hot natures and polysaccharides and find one suitable statistical pattern recognition for explaining this correlation.Methods： Extracted polysaccharides from 60 kinds of traditional Chinese medicines（TCM）.The polysaccharides were completely hydrolyzed into monosaccharides.After that,the obtained derivatives from monosaccharide were separated by high performance liquid chromatography（HPLC）,and constructed the database.Based on HPLC data,taking ＇exhaustive attack＇ and ＇complete trial-and-error＇ method as model selected strategy combined with statistical simulation,and compared the performance of six well-known classification methods which could be reasonably easily adapted to the analysis of HPLC data.Results： Statistical simulation showed that PLS-DA model had higher sensitivity for CHMP-markers recognition than the other models.In PLS-DA model,the forecast accuracy was 91.7%.Conclusion： There was a significant correlation between Cold-Hot natures and polysaccharides,and PLS-DA model was determined as the most appropriate pattern recognition model for CHMP-markers from so many models.