서강대학교

목차

Abstract (English) 8

Abstract (Korean) 11

Ⅰ. Introduction 14
1.1 Background 14
1.2 Authenticity issues in edible oils and sesame oil 15
1.3 Current analytical approaches and limitations 17
1.4 Aim of the study and research questions 19
1.5 Overview of the proposed methodological framework 21
1.6 Structure of the thesis. 23

Ⅱ. Theoretical Background 25
2.1 Chemical composition of edible oils 25
2.1.1 Fatty acids in edible oils 25
2.1.2 Phytosterols and other minor components 26
2.2 GC–MS and GC–MS/MS for compositional analysis 28
2.3 LC–MS/MS and data-dependent MS/MS (DDA) 29
2.4 Multivariate analysis for authenticity assessment 31
2.4.1 Principal Component Analysis (PCA) 31
2.4.2 Partial Least Squares–Discriminant Analysis (PLS–DA) 32
2.5 Machine learning models for classification and regression 33
2.5.1 Support Vector Machine (SVM) 33
2.5.2 Performance metrics 34
2.6 Spectral library search and MS/MS molecular networking 36
2.6.1 Mass spectral libraries and search strategies 36
2.6.2 Cosine similarity and mirror plots 37
2.6.3 MS/MS molecular networking 38

Ⅲ. Materials and Methods 40
3.1 Samples and blending design 40
3.1.1 Reference oils and commercial sesame oil samples 40
3.1.2 Blending scenarios and mixing ratios for model samples 42
3.2 Sample preparation 44
3.2.1 Preparation of fatty acid methyl esters (FAMEs) for
GC–MS 44
3.2.2 Extraction and derivatization of sterols and metabolites 45
3.3 Instrumental conditions 46
3.3.1 GC–MS method for fatty acid analysis 46
3.3.2 GC–MS/MS method for sterol and metabolite analysis 47
3.4 Quantification and data pre-processing 48
3.4.1 Calibration, LOD/LOQ, linearity and RSD 48
3.4.2 Peak detection, alignment and normalization 49
3.4.3 Data filtering and feature selection 50
3.5 Multivariate analysis and SVM modelling 51
3.5.1 Data splitting and cross-validation strategy 51
3.5.2 SVM classification for adulteration detection 52
3.5.3 SVM regression for mixing ratio estimation 53
3.6 Design and implementation of the molecular identification and networking software 54
3.6.1 Overall architecture and pipeline design 54
3.6.2 GC–MS and LC–MS/MS processing modules 55
3.6.3 Spectral library formats and search algorithms 56
3.6.4 Graphical User Interface (GUI) and user workflow 57
3.7 Multi-layer authenticity assessment framework 58
3.7.1 Concept and role of each analytical layer 58
3.7.2 Criteria for combining analytical evidence 59

Ⅳ. Results and Discussion 62
4.1 Quantitative performance of GC–MS and GC–MS/MS methods 62
4.1.1 Linearity, LOD/LOQ and precision for fatty acids 62
4.1.2 Quantitative characteristics of sterol and metabolite
analysis 63
4.2 Fatty acid compositional profiles of oils and blends 64
4.2.1 Fatty acid composition of reference oils 64
4.2.2 Changes in fatty acid profiles according to mixing ratio 66
4.3 Sterol and metabolite profiles 69
4.3.1 Sterol composition across oils and extraction solvents 70
4.3.2 Metabolite features and candidate markers 71
4.4 Multivariate analysis of compositional data 73
4.4.1 PCA/PLS–DA of fatty acid data 73
4.4.2 PCA/PLS–DA of sterol/metabolite data 75
4.4.3 Integrated interpretation of lipidomic layers 80
4.5 Performance of SVM-based authenticity models 81
4.5.1 Classification of oil type and adulteration status 81
4.5.2 Estimation of mixing ratios and prediction error 83
4.5.3 Strengths and limitations of SVM models 84
4.6 Molecular identification and networking software 85
4.6.1 GC–MS-based identification of marker compounds 85
4.7 Proposal of a multi-layer authenticity assessment workflow 91
4.7.1 Role of GC–MS, sterol/metabolite analysis and LC– MS/MS networking 91
4.7.2 Example decision scenarios for sesame oil authenticity 93
4.7.3 Practical considerations for regulatory and industrial application 94

Ⅴ. Conclusion 96
5.1 Summary of major findings 96
5.2 Methodological contributions of this study 98
5.3 Limitations and future perspectives 101
References 104

List of Tables and Figures
Figure 1 Comparison of major fatty acid compositions in four reference oils. Values represent mean percentages of total fatty acids for sesame, soybean, canola and corn oils. 66
Figure 2 Heatmap of fatty acid compositions for external blind blends of sesame oil with soybean, canola and corn oils. Each row represents a blind sample, and each column
represents a fatty acid expressed as percentage of total fatty acids 68
Figure 3 PCA/PLS-DA of fatty acid data 74
Figure 4 PCA and PLS-DA of sterol and metabolite profiles obtained from n-hexane extracts 76
Figure 5 PCA and PLS-DA of sterol and metabolite profiles obtained from ether extracts 77
Figure 6 Integrated PCA and PLS-DA combining fatty acid, sterol and metabolite variables 79
Figure 7 Performance of SVM-based authenticity models using fatty-acid profiles 82
Figure 8 User interface of the NetMol Insight software 87
Figure 9 Overall layout of the NetMol Insight graphical user interface 88
Figure 10. Molecular identification interface of NetMol Insight . 89
Figure 11 MS/MS molecular networking view in NetMol Insight 90
Figure 12 Multi-layer workflow for edible oil authenticity assessment proposed in this study 92

Table 1 Summary of samples and data scales used in this study 41
Table 2 Design of blind adulteration scenarios and mixing series 43
Table 3 The detailed compositions of four reference oils are summarized in Table 65

Abstract (English)

Abstract (Korean)