Studies On Similarity Evaluation Of Chromatographic Fingerprint For TCM

Part IChromatographic fingerprinting combined with similarity evaluation provides an effective solution for the modernization of TCM (traditional Chinese medicine) quality control at present. Our study ranges from the theoretical induction to the software programming, and then the experimental verification.Effects of different retention time correction formulae on linear shift, nonlinear shift and retention time delay are discussed. Single-point formula deals well only with linear shift. Dual-point formula is suitable for the correction of nonlinear shift and retention time delay. Multi-point formula, depending on the segmented linear approximation, turns out to be effective in correcting the nonlinear shift.Peak matching is one of the key steps in the software programming, which improve the reliability of similarity calculation. A new matching algorithm is put forwarded based on total peaks sorting.Sensitivity and stability of three formulae in calculating similarity, i.e., cosine, correlation coefficient and Euclidean distance, are compared. Sensitivity and stability of the similarity, calculated by cosine with the point-to-point and peak-to-peak data, are further contrasted through derivation. Cosine with the peak-to-peak data provides an effective approach resulting in a compromise between sensitivity and stability.Accordingly, a chromatographic fingerprinting workshop is developed. The software provides the function of similarity calculation, integrated with three pattern recognition modules, i.e., principle component analysis, cluster analysis and artificial neural network.