基于聚类、主成分和判别分析的海马醇提物红外指纹图谱研究

Infrared Fingerprint Analysis of Hippocampus′s Ethanol Extracts Coupled with Cluster Analysis,Principal Component Analysis and Discriminant Analysis

目的利用红外光谱指纹图谱技术和数理统计分析方法建立一种海马(Hippocampus)种类鉴别和质量控制方法。方法应用红外光谱技术分析35个不同来源的海马样品乙醇提取物,通过聚类分析和相似度分析建立海马乙醇提取物的红外光谱指纹图谱;利用主成分分析及判别分析进行海马种类鉴别和质量控制研究。结果各海马样品乙醇提取物的红外光谱与所建立的指纹图谱共有模式比较相似度良好,符合指纹图谱要求;三斑海马(Hippocampus trimaculatus)与刺海马(H.histrix)分别聚集为不同的类群,实现了不同种类海马的鉴别;在主成分分析的基础上进行判别分析时,吸收峰3 305.4～3 297.7 cm-1、2 919.7～2 915.9 cm-1、2 852.2～2 848.4 cm-1、1 656.6～1 652.7 cm-1、1 542.8～1 538.9 cm-1、1 456.0～1 452.2 cm-1在主成分1上的载荷较大,而1 045.2～1 041.4 cm-1在主成分2上的载荷最大,据此建立的判别函数可用于区分三斑海马与刺海马,正确率达到93.55%。结论本实验建立的乙醇提取物红外光谱指纹图谱结合主成分分析及判别分析可快速、准确地鉴别不同种类的海马。

To establish and validate an infrared spectrometry method coupled with cluster analysis, principal component analysis and discriminant analysis for identifying different kinds of Hippocampus. METHODS Fourier transform infrared （ FrIR） spectrometer was used to analyze the ethanol extracts from 35 samples of Hippocampus collected from different geographical loca tions. Similarity analysis and cluster analysis were used to establish the reference fingerprint spectra, and principal component analysis and discriminant analysis were used for identifying different kinds of Hippocampus. RESULTS The samples of the same species had similar characteristic peaks of infrared spectra and could be clustered together. Based on the result of principal component analysis, the absorption peaks （3 305.4 -3 297.7 cm-1 , 2 919. 7 -2 915.9 cm-1 , 2 852. 2 -2 848.4 cm-1 , 1 656. 6 - 1 652.7 cm-l , 1 542. 8 - 1 538.9 cm-1 , 1 456. 0 - 1 452. 2 cm-1 and 1 045.2 - 1 041.4 cm-1 ） were the main characteristic peaks and were used to develop the discrimination function for distinguishing different species of Hippocampus. The accuracy of the discrimination function was 93. 55%. The established method had good stability, precision and reproducibility, and was a very reliable and useful method for dis- tinguishing different species and assessment of the quality of Hippocampus. CONCLUSION The established IR fingerprints of Hippocampus＇s ethanol extracts coupled with mathematical statistics theory can be applied to identify different kinds of Hippocampus. More importantly, the research has provided a new thought for traditional Chinese medicine identification by using mathematical model. KEY WORDS ： Hippocampus ; ethanol extract; infrared spectrum; fingerprint