The New Applications Of Tchebichef Image Moments In Analysis

In recent years,the developments of instrumental chemical analysis techniques and computational chemistry have generated large multidimensional data which contains more comprehensive chemical information,but at the same time increased the difficulties of extracting effective features and accurate multidimensional calibration.As a technique of feature extraction,image moment method has been applied to some three-dimensional spectra analysis in the chemical field.In order to further explore the universality and stability of image moments and extend more practical application,it must be explored that feasibility and advantages of image moment method in new application area of analytical chemistry.As one of discrete orthogonal moments,Tchebichef moments have excellent performance in image moments.In addition to translation,rotation,scaling invariance and multiresolution capabilities,data redundancy and discretization errors are also reduced in calculation of Tchebichef moments which make them more accurate and robust.Therefore,we applied Tchebichef moments to quantitative analysis of two types of data in this study.One is the three-dimensional spectral data in the full scan mode of high performance liquid chromatography-mass spectrometry(HPLC-MS).Unlike the spectrogram or chromatogram,the signal in mass spectrogram is rod-shaped discrete.Although its full scan mode can be used to monitor all the signals simultaneously in a large range,it is difficult to be applied to quantitative analysis of complex system because of the overlapping peaks,unknown signal interference and other problems.The other is molecular structure two-dimensional image data,based on which quantitative structure-activity relationship(QSAR)was analyzed.This new developed QSAR approach uses multivariate image analysis based on the chemical structure itself rather than complicated quantum descriptors.But the biggest drawback in this method is the requirement for pixel alignment of parent chemical structure,which need a lot of work in QSAR study with many samples.Therefore,we apply the mature Tchebichef moments to above two types of data to solve the practical problems in analytical chemistry.At the same time,the Tchebichef moment method is rigorously verified to ensure the credibility of the results and the superiority of the method.The paper mainly includes:In Chapter 1,the background,development status and prospect of chemometrics and image moments are briefly summarized,and the main research contents of this dissertation are presented.In Chapter 2,Tchebichef moment method was applied to quantitative analysis of geniposide,forsythin and gentiopicroside in Qingrejiedu oral liquid based on HPLC-MS full scan mode.Tchebichef moments were extracted directly from the three-dimensional spectrum,and quantitative linear models of three active compounds were established by stepwise regression algorithm.All correlation coefficients were more than 0.9978.The limits of detection and the limits of quantification were less than 0.11 μg mL-1 and 0.49 μg m L-1,respectively.The intra-day and inter-day precision,calculated as relative standard deviation(RSD)were less than 6.54% and 9.47%,while the recovery ranged from 102.56% to 112.15%.Owing to the advantages of multi-resolution and inherent invariance properties in translation,rotation and scaling,Tchebichef moment method could provide favorable results even in the situation of peaks shifting and overlapping,unknown interferences and noise signals.In Chapter 3,molecular structure images were combined with Tchebichef moment method for quantitative structure-activity relationship analysis of anti-HIV-1 compounds.For a series of HIV-1 non-nucleoside reverse transcriptase inhibitors,Tchebichef moments extracted from molecular structure images were utilized as the descriptors of chemical information and employed to establish the linear QSAR models through two simple regressions algorithms,partial least squares(PLS)and stepwise regressions.In order to verify the reliability of the method,we have verified the model sufficiently and strictly.It is proved that the Tchebichef moment method can effectively extract the comprehensive chemical characteristics in molecular structure images without any image preprocessing,and can obtain the satisfactory result.In Chapter 4,molecular structure images were combined with Tchebichef moment method for toxicity assessment of anilines and phenols to aquatic organisms.Environmental risk assessments of anilines and phenols,two mass-produced industrial chemicals,have great realistic meaning.Based on molecular structure images,Tchebichef moments were employed to build the quantitative structure-activity relationship(QSAR)models for the toxicity predictions of anilines and phenols towards two aquatic organisms,algae P.subcapitata and bacteria V.fischeri.Compared with published works,our models possess higher performance and predictive capability,and can avoid the pretreatment of image data.Therefore,the proposed method provides a new strategy for the QSAR study to apply to the environmental risk assessment of toxic compounds.Chapter 5 includes main conclusions and outlook.The advantages of the Tchebichef moment method in the above two kinds of data are summarized,and the application of the Tchebichef moment method in the forefront of analytical chemistry is prospected.