검색 상세

Applicability of fluorescent sensor arrays using kaleidoscopic indolizine for various analytes based on machine learning

  • 김현기 (아주대학교 일반대학원)

초록/요약

As time has passed, humanity's interest in health has increased. Prior to treating a disease, it is often the aim of medics to identify the disease, as is currently the case with COVID-19. We suggest that the single most important factor is the cause of the disease to be treated. Consequently, we aimed to develop a simple and noninvasive sensor that could monitor health. We selected a fluorescent material that mimicked the function of mammalian olfactory receptors and simultaneously acted as a signal element. Fluorescent compounds based on indolizine structures are known to exhibit an existing intramolecular charge transfer (ICT) phenomenon, and it was expected that their photophysical properties would change due to electron density by interaction with the target analytes [1]. Our team have been able to develop numerous fluorescence frameworks through combinatorial chemical synthesis and have confirmed the emission of various fluorescent materials in the liquid and solid states at a single excitation wavelength (ex 365 nm) [2]. We term this the Kaleidoscopic Indolizine (KIz) system. One KIz fluorescent compound reacts with the analyte to produce a color difference value in response to exposure (before and after). Multiple designs of these fluorescent compounds form a specific pattern for the analyte, while forming an array of fluorescent sensors. Conventional array-type sensing systems require specialized operators with technical knowledge to measure the accuracy of electrochemical systems. To overcome this, we introduced an image analysis system that can detect various analytes using globally available mobile devices. A sensor array composed of multiple fluorescent compounds reacts with the analyte to produce a unique pattern. A highly accurate pattern-recognition machine learning algorithm was introduced into our sensing system. Pattern recognition and machine learning are commonly integrated fields of artificial intelligence (A.I). As our novel fluorescent sensor array is supported by machine learning, inexperienced users with no expertise can use it as a sensor with high accuracy. As a result, we developed a fluorescent sensor array that uses non-invasive methods to monitor health, and produces specific, identifiable patterns for volatile organic compounds, pH, and glucose. Based on the results of developing a sensor platform that can detect a variety of analytes using a chemical approach, if a bioengineering approach is adopted, it is expected that a diagnostic platform with higher performance will be built.

more

목차

Chapter I. Introduction 1
1.1 Electronic nose sensor 2
1.2 Colorimetric sensor array 4
1.3 Fluorescent sensor array 4
1.3.1 Volatile organic compounds 5
1.3.2 Heavy metal ion 8
1.3.3 pH 8
1.3.4 Pathogen 9
1.3.5 Saccharide 12
1.4 Random Forest (Machine learning) 12
1.5 Objectives of a thesis 15
1.6 Reference 17

Chapter II. Kaleidoscopic fluorescent arrays for machine learning-based point-of-care chemical sensing 27
Abstract 28
2.1 Introduction 29
2.2 Experimental section 31
2.2.1 Materials and compound characterization 31
2.2.2 Absorption and fluorescence related properties 31
2.2.3 Solid state photophysical property measurement 32
2.2.4 Quantum mechanical calculations 32
2.2.5 Experimental procedure for sensor array preparation by wax printing 32
2.2.6 Smartphone-based detection system 32
2.2.7 Automated color information extraction from captured fluorescent array images 33
2.2.8 VOC classification with machine learning model 34
2.3 Result and discussion 34
2.3.1 Design of fluorescent compounds and detection systems 34
2.3.2 Photophysical properties of KIz compounds 38
2.3.3 Environmental chemical sensing with KIz compounds 41
2.3.4 VOC classification with KIz compound arrays 41
2.3.5 Machine learning-based data analysis 44
2.4 Conclusion 47
2.5 References 49

Chapter III. Fluorescent sensor array for high-precision pH classification with machine learning-supported mobile devices 53
Abstract 54
3.1 Introduction 54
3.2 Experimental section 57
3.2.1 Materials and compound characterization 57
3.2.2 Photophysical properties 57
3.2.3 Experimental procedure for sensor array preparation with wax printing 57
3.2.4 Sensor array and pH solution exposure method 58
3.2.5 Smartphone-based detection system 58
3.2.6 Automated color-information-extraction from the captured fluorescent array images 58
3.2.7 pH discrimination with a machine learning model 59
3.3 Result and discussion 59
3.3.1 Design and synthesis of 30 different fluorescent probes for accurate pH classification 59
3.3.2 Spectral and photophysical properties of the 30 different compounds 64
3.3.3 pH-Responsive behavior of fluorescent probes 67
3.3.4 Fluorescent sensor array preparation 70
3.3.5 Extracting the ROI region and features 72
3.3.6 Model design 76
3.4 Conclusions 79
3.5 References 80

Chapter IV. Fluorescent sensor array based indolizine for noninvasive monitoring tear glucose with machine learning-supported smartphone 85
Abstract 86
4.1 Introduction 88
4.2 Experimental section 89
4.2.1 Compound characterization 89
4.2.2 Materials 89
4.2.3 Photophysical properties study 90
4.2.4 Preparation of analytes solution 90
4.2.5 Experimental Procedure for sensor array preparation with wax printing 90
4.2.6 Preparation of artificial tear fluid 91
4.2.7 Tear glucose solution exposure on sensor array method 91
4.2.8 Smartphone-based analyzer system 91
4.3 Result and discussion 93
4.3.1 Molecular design of four different fluorescent biosensors 93
4.3.2 Spectral and photophysical properties of 4 different biosensors 94
4.3.3 Effect of H2O2 on photophysical property of AFGlu 1–4 94
4.3.4 Selectivity of AFGlu 1–4 against H2O2 95
4.3.5 Fluorescent sensor array preparation and automated color feature extraction from fluorescent array images 98
4.3.6 Fluorescent sensor array exhibiting unique fluorescent pattern selectively against D-glucose 101
4.3.7 Model design and feature engineering 101
4.3.8 Model performance evaluation 105
4.4 Conclusions 106
4.5 References 109

Chapter V. Conclusions & Perspectives 115
5.1 Conclusions 116
5.2 Perspectives 116
Appendix: Chapter II 117
Appendix: Chapter III 357
Appendix: Chapter IV 484

more
반출 Meta View 목록