- 學位論文
光致發光奈米材料與深度學習應用於濫用藥物快篩
Application of Photoluminescent Nanomaterials and Deep Learning for Abused Drugs Screening
摘要
我們的研究目標係利用新穎奈米材料發展毒品快篩試劑，以克服現行化學呈色試劑及免疫法檢測試紙對於常見及新興毒品選擇性不佳或廣篩能力不足之缺點。研究開端係由L-精胺酸通過水熱法製備的碳點（C-dots）及其碳點功能化紙（C-dot-functionalized papers，CDFPs），用於定量水溶液中4-氯乙基卡西酮。為製備CDFP，利用固態墨水印刷法製作每張含8×12圓孔圖案的層析紙，然後將碳點塗到圓孔中。π共軛的酮或酯化合物通過電子轉移機制誘導碳點的光致發光（PL）淬息。CDFP可以篩驗古柯鹼、海洛因和卡西酮等濫用藥物。由於在pH 11.0下海洛因和古柯鹼的溶解性較差，因此碳點探針對卡西酮具有選擇性。碳點於水溶液中及CDFP上在pH 11.0條件下對4-氯乙基卡西酮之最低偵測極限分別至1.73 mM和0.14 mM。我們的感測系統由攜帶式紫外線燈、智能手機和低成本CDFP組成，可用於檢測卡西酮、古柯鹼及海洛因，顯示在犯罪現場篩選這些濫用藥物的潛力。 第二部分中，疏水性碳點（hC-dots）係通過水熱法由D-苯丙胺酸製備，用於快篩飲料中濫用的強姦藥物，如硝甲西泮、氟硝西泮、氯硝西泮及硝西泮。為減少各種飲料（包括啤酒，紅酒，威士忌和橙汁）的基質干擾，當以波長365 nm激發進行分析物誘導hC-dots螢光於430nm淬息定量前，需使用甲苯進行液‒液萃取分析物5分鐘。本檢驗法可對硝甲西泮定量之濃度可達7.24 μM，低於飲料中藥物輔助性侵之濃度。本檢驗方法對硝基取代之苯二氮平類藥物具有選擇性，而非常見的濫用藥物，包括4-氯乙基卡西酮、古柯鹼、海洛因、4-羥基丁酸、愷他命及甲基安非他命等。因僅需要hC-dots（分散在0.5 mL甲苯中）、手持式紫外光（365 nm）及照相機（或智能手機），這種簡單、低成本、快速且具選擇性的感測系統非常適合用於藥物輔助性侵的犯罪現場進行飲料篩驗。 最後一部分則係利用金/銀奈米團簇、碳量子點及經Marquis reagent反應後產生之螢光聚合物粒子整合成感測陣列，並結合卷積神經網絡之深度學習工作站發展深度學習毒品篩驗平台系統（drug screening platform and system，DL-DSPS），可透過遠端遙控即時檢測古柯鹼、海洛因、甲基安非他命、愷他命及合成卡西酮。這些濫用藥物通過電子轉移機制或聚集誘導發射及產生螢光聚合物粒子等，誘導奈米材料PL淬息或增強以區別濫用藥物。高至低濃度之愷他命、海洛因、甲基安非他命、4-氯乙基卡西酮、古柯鹼及葡萄糖樣品的陣列PL圖像（共1,230張照片）用於訓練和優化該DL-DSPS。訓練完成後，另一來源的PL陣列圖像（共1,560張照片）用以驗證DL-DSPS。當愷他命、海洛因、甲基安非他命、4-氯乙基卡西酮及古柯鹼的濃度分別高於0.8、2、2、2及1.5 mM時，DL-DSPS檢測準確度達到100%。此外，所開發的DL-DSPS 也以真實非法藥物進行驗證，篩驗結果與GC-MS鑑定無顯著性差異，顯示該平台系統於犯罪現場進行多種藥物篩驗之潛力。 本研究證明了奈米材料用於藥物快篩係可行且值得開發，可克服化學呈色測試和免疫分析試劑組的不足。
並列摘要
The aim of our work is to develop drug-screening reagents by using novel nanomaterials, to overcome for the shortcomings of current chemical color reagents and immunoassay test strips, that suffering poor selectivity or insufficient broad screening capabilities for common and emerging drugs. First of all, carbon dots (C-dots) prepared from L-arginine through a hydrothermal route and their C-dot-functionalized papers (CDFPs), have been used for quantitation of 4-chloroethcathinone in aqueous solution. To prepare CDFPs, chromatography papers, each with a pattern of 8 × 12 circles (wells), are fabricated through a solid-ink printing method and then the C-dots are coated into the wells. π-Conjugated keto or ester compounds induce photoluminescence (PL) quenching of C-dots through an electron transfer process. The CDFPs allow screening of abused drugs such as cocaine, heroin and cathinones. Because of poor solubility of heroin and cocaine at pH 11.0, the C-dot probe is selective for cathinones. The C-dots in aqueous solution and CDFPs at pH 11.0 allow detection of limits for 4-chloroethcathinone down to 1.73 mM and 0.14 mM, respectively. Our sensing system consisting of a portable UV-lamp, a smartphone, and a low-cost CDFP has been used to detect cathinones, cocaine and heroin, showing its potential for screening of these drugs in crime sites. In the second part, hydrophobic carbon dots (hC-dots) are prepared from D-phenylalanine through a hydrothermal route, which have been used for screening of abused date rape drugs in beverages such as nimetazepam, flunitrazepam, clonazepam, and nitrazepam. To minimize matrix interference in various beverages, including beer, red wine, whisky, and orange juice, liquid-liquid extraction of the analytes using toluene for 5 min is required before quantitation based on analyte induced fluorescence quenching of the hC-dots at 430 nm when excited at 365 nm. This assay allows the quantitation of nimetazepam down to 7.24 M, which is lower than that found in the beverages associated with drug-facilitated sexual assault (DFSA). The assay is selective toward nitro-substituted benzodiazepines over popular abused drugs, including 4-chloroethcathinone, cocaine, heroin, 4-hydroxybutyric acid, ketamine, and methamphetamine. Since only hC-dots (dispersed in 0.5 mL toluene), a handheld UV light (365 nm), and a camera (or smartphone) are required, this simple, low-cost, rapid, and selective sensing system is ideal for beverage analyses at DFSA crime scenes. In the last part, gold / silver nanoclusters, C-dots, and fluorescence polymer particles (FPPs) produced after reactions with Marquis reagent are integrated as a sensing array, that combined with a convolutional neural networked (CNN) of deep learning workstation to develop a drug screening platform and system (DL-DSPS) for detection of cocaine, heroin, methamphetamine, ketamine, and synthetic cathinones, through real-time access and support. These abused drugs induced PL quenching or enhancing of nanomaterials through electron transfer mechanism or aggregation-induced-emission (AIE), as well as the formation of FPPs for discrimination of abused drugs. The arrayed PL images (1,230 pictures) from real samples of ketamine, heroin, MA, 4-CEC, cocaine, or glucose with high to low concentrations are utilized to train and optimize DL-DSPS. After training completed, the arrayed PL images (total 1,560 pics) from another source are performed to validate the DL-DSPS. The 100% accuracy is reached for detection of ketamine, heroin, MA, 4-CEC, cocaine, when its concentration is higher than 0.8, 2, 2, 2, and 1.5 mM for ketamine, heroin, MA, 4-CEC, and cocaine, respectively. In addition, the developed DL-DSPS has been validated for screening of real-world illicit drug samples, with results in good agreement with that from GC‒MS, showing its potential for multi-drug screening at crime scenes. Our studies demonstrated nanomaterials used for drug screening is practical and worthy for development to overcome the insufficiency of chemical color tests and immunoassay kits.