An Improvement Of The Calibration Results For Grey Analytical System In HPLC Method Applying CBBL Algorithm Based On GA Optimization Strategy

HPLC has a wide range of applications in chemistry, biology, clinic, pharmacy, food and environment monitoring fields. It has become an indispensable tool for chemical separation and detection. With the wide applications of HPLC separation technique, the mixed samples are always found containing unknown background components with overlapping peaks. It can’t always rely on one-dimensional data to achieve successful quantitative and qualitative analysis. HPLC-DAD can provide two-dimensional data including more useful chemical information. It can provide chemists some new methods to calibrate the grey analytical system (The so-called grey analytical system is between white and black analytical system. The target components exist in the analytical sample, but it is not clear whether there is another unknown interference or background[1]. The analysis of the grey system has two methods, which includeVectorial Calibration Methods (VCM) and Matrix Calibration Methods(MCM). Vectorial Calibration Methods have three main methods,which include Additional Iterative Target Transformation FactorAnalysis (AITTFA), Adaptive Kalman Filtering (AKF), and Local CurveFitting (LCF). Through the combination of standard addition、matrix projection algorithm and iterative transformation method,the unknown background can be detected applying AITTFA, and then,the quantitative calibration of the grey system can be realized【2】. Torealize the resolution for the grey analytical system, AKF utilizes thecharacteristics of the innovation sequence and adjusts the errors ofthe measurements of the corresponding data points to resist theerrors of the models【3】. The main ideas of LCF applied in quantitativecalibration for grey system are that the values of the differentiatedbackground spectra are zero at the maximum value points of theoriginal background spectra【4】. However, Vectorial CalibrationMethods require strong prerequisites, and the degree of overlapbetween the background interference signal and the measuredcomponent signal is very weak or it only has the measuredcomponent signal in a certain wavelength range. Generally VCA onlygives possible solution and poor practical performance[5]. MCM is ahot chemometric research in recent years, which include Generalized Rank Annihilation Factor Analysis (GRAFA)[6-7], ResidualBilinearzation Method (RBM)[8], and Constrained BackgroundBilinearzation Method(CBBL)[9]. The unique solutions in physicalsense can be achieved by these methods. However, when HPLC greysystem was processed by Matrix Calibration Methods, thequantitative calibration results are still poor if the relevantcomponents have poor reproducibility in retention time[9]. Aiming atthis problem, firstly, the matrix data in this paper was de-noised byWavelet Transform method. Then the response matrix ofchromatographic system with unknown interference background wasdecomposed by CBBL method. Consequently, the retention time andthe concentrations of determining components in the greychromatographic system were optimized and quantifiedsimultaneously by GA. In the case of poor reproducibility in theretention time of relevant components in HPLC methods, theaccuracy of calibration was improved by the method proposed in thisdissertation.The main work of this dissertation:(1)The chemometric methods applied in the dissertation weredescribed briefly, which include constrained backgroundbilinearization method (CBBL), genetic algorithm (GA), and wavelettransform (WT). (2)Firstly, two-dimensional bilinear data of multicomponentoverlapping chromatographic peaks was simulated by EMG(Exponentially Modified Gaussian, EMG), and the two-dimensionalbilinear data was de-noised by WT. Then, according to theproportional relationship among the concentrations of determiningcomponents (the determining and the standard samples contain thesame components, but the retention time of the determiningcomponents shift randomly.), the peak areas and the responsematrix, the real concentrations of the determining componentsin the grey analytical system were calculated. Finally, the simulatedresponse matrix of two-dimensional bilinear chromatographic systemwas decomposed by CBBL method, and the background matrix wasrebuilt by PCA. Consequently, the retention time and theconcentrations of the determining components were optimized andquantified by GA simultaneously. The calculated results of thesimulated data show that even the retention time of relevantcomponents has poor reproducibility, the concentrations and theretention time can be simultaneously and accurately quantified. Thecalibration accuracy was improved by the method proposed in thisdissertation.(3)The two-dimensional data of four-component and six–component samples were collected from HPLC experiments, which were analyzed by the proposed method in the dissertation. Firstly,WT method was used to de-noise the two-dimensional bilinearchromatographic data. Then, the response matrix of two-dimensionalbilinear grey chromatographic system was decomposed by CBBLmethod, and the concentrations and the retention time of thedetermining components were optimized and quantified by GAsimultaneously. The calculation results of the experimental datashow that even the relevant components have poor reproducibility inretention time, the concentrations and the retention time of therelevant components can also be simultaneously and accuratelyquantified. The calibration accuracy of the actual chromatographicsystem was improved by the method proposed in this dissertation.