Deep Learning-assisted Three-dimensional Fluorescence Difference Spectroscopy For Rapid Identification And Quantification Of Illicit Drugs In Bio-fluids

In this thesis,common-used analysis methods for illicit drugs,recent application of deep learning in chemistry and the concept and principles of fluorescence analysis were reviewed.The fast identification and quantification of illicit drugs in bio-fluids are of great significance in clinic detection.However,existing drug detection strategies cannot fully meet clinic needs and the on-site identification and quantification of various illicit drugs in bio-fluids remains a great challenge.The emergence of artificial-intelligence(AI)technologies provide n ew avenues to model complex scientific problems even without knowing clear mechanism.Fluorescence analysis has been one of the most typical methods in analytical chemistry,however,for information-rich three-dimensional(3D)fluorescence spectra,current analyzing methods still need to be improved to dig more information from it.In this thesis,we proposed a deep learning-assisted three-dimensional fluorescence difference spectroscopy for rapid identification and quantification of illicit drugs in bio-fluids.With silver nanoclusters(AgNCs)as signal sources,the interaction between AgNCs and drug molecules led to a change in fluorescence performance,and deep learning methods were applied to extract the subtle fingerprint information from the difference spectra to identify and quantify various illicit drugs in bio-fluids,including codeine,4,5-methylene-dioxy amphetamine,3,4-methylene dioxy methamphetamine,meperidine and methcathinone.The main contents are summarized as follows:The AgNCs with an average diameter of about 2 nm were prepared by reducing aqueous AgNO3 with glutathione under vigorous stirring in an alkaline solution.The as-prepared nanoclusters have good fluorescence performance with two emission peaks at emission wavelengths of 380 nm and 430 nm,respectively.The highly fluorescent AgNCs acting as a strong signal source were added into urine containing drugs.Every individual illicit drug could produce a distinct output signal in the difference spectrum,which might be related to the differences in the functional groups of the drug molecules.After model optimization among 6 classifiers(including generative adversarial network,artificial neural network,K nearest neighbors,naive bayes,support vector machine and convolutional neural network),generative adversarial network was chosen to classify drugs in human urine and for each king of drug,artificial neural network was used to quantify the specific drug.The strategy is able to identify drugs in human urine with a high prediction accuracy rate of 88.74%and quantify drugs with a broad detection range from 2?g/ml to 100 mg/ml,which are comparable with those obtained by liquid chromatography-mass spectrometry.The proposed strategy is also rapid,simple,low-cost,recognition element-free and separation-free.Thus,this work provides a novel method for rapid scan of illicit drugs in suspected urine samples in forensic area.The combination of deep learning with the information-rich 3D fluorescence difference spectrum derived from subtle interaction between fluorescent AgNCs and small molecule drugs provides a promising tool to identify and quantify illicit drugs in bio-fluids,even without knowing the exact interaction mechanism.It opens up a new way for the detection of small molecules with or without fluorescence in complicated matrices.