Terahertz Spectroscopic Analysis Of Single-component And Multi-component Substances Based On Support Vector Regression

In this dissertation,the quantitative and qualitative analysis of single-component bovine serum albumin and multi-component environmental hormone bisphenol derivatives were performed by the support vector regression machine.Secondly,the principle of the support vector regression(SVR),maximum information coefficient(MIC)and terahertz spectral pre-processing methods are introduced in detail,and a terahertz spectral feature selection and quantitative analysis software is designed and used for data analysis of related experiments in this paper.The core content of this paper includes:First,the material detection analysis of single-component biofilms based on support vector regression was performed.Quantitative and qualitative analysis of bovine serum albumin(BSA)deposited films was achieved by combining machine learning methods with terahertz time-domain spectroscopy.The basic model for the relationship between target concentration and frequency was established using a support vector regression machine for the deposition of thin films of BSA(21 different concentrations,147 samples)in the concentration range of 0.5-35 mg/ml,and a successful prediction of the unknown sample concentration was achieved with a determination factor of R~2=0.97932.In addition,the correlation between each frequency and concentration was identified using the Maximum Information Coefficient(MIC)method.The theoretical calculations obtained three maximum distinguishable frequencies in the terahertz band(1.198,1.078,and 0.479 THz),and it was determined that these three characteristic frequencies correspond to the fundamental frequency vibrational frequency of the long-wavelength elastic vibration model of the BSA protein.Second,the environmental hormones in multi-component complex systems detection analysis based on support vector regression was performed.We used the environmental hormone bisphenol A(BPA)as the main research object,performed qualitative analysis of three derivatives with similar molecular structures(BPAF,BPE,and BPS),analyzed the vibration-related frequencies of the four substances using principal component analysis(PCA),and determined two sets of relevant frequencies:(1)2.20,2.43,2.49,and 2.52 THz;(2)1.64,1.70,1.82,and 1.85 THz,demonstrated the similarity of the properties of the four bisphenols.In addition,the concentration of BPA was predicted for different concentrations of bisphenol mixtures(37 different concentrations,148 samples).The prediction coefficients of determination were R~2=0.99266,0.98889,0.9836,and 0.98001,respectively.The terahertz absorption spectra were inversely predicted and the results agree well with the measured spectral curves.Finally,a graphene oxide composite metal ion(Go/Metl-NPs)microfluidic chip based on transparent plastic substrate was designed.The surface of the chip was modified with graphene oxide(Go)and metal nanoparticles(Metl-NPs).The large specific surface area of Go was used to load more Metl-NPs and to adsorb partial proteins,Metl-NPs adsorbed proteins through physical interactions.Experiments have shown that the complex can effectively enhance the THz signal,and the study will prepare for the quantitative and qualitative analysis of cancer tumor marker proteins to be developed in the future.