應(yīng)用紅外光譜和化學(xué)計(jì)量學(xué)進(jìn)行疾病診斷及甘草指標(biāo)成分含量測定的研究
發(fā)布時(shí)間：2018-08-18 14:43
【摘要】：本論文主要研究建立基于衰減全反射傅里葉紅外光譜(Fourier Transform Infrared Spectroscopy-Attenuated Total Refraction,FTIR-ATR)和近紅外光譜的定性和定量模型,用于快速篩查新生兒苯丙酮尿癥(Phenylketonuria,PKU)及同時(shí)測定甘草飲片中甘草苷和甘草酸兩種指標(biāo)成分的含量。研究采用了平滑、求導(dǎo)、主成分分析、無信息變量消除、歸一化等預(yù)處理方法,并以偏最小二乘法、多模型共識偏最小二乘法(consensus Partial Least Squares,cPLS)、核偏最小二乘法(Kernel Partial Least Squares)、多模型共識核偏最小二乘法(consensus Kernel Partial Least Squares)建立了目標(biāo)成分含量的定量校正模型,經(jīng)多種評價(jià)指標(biāo)對模型性能進(jìn)行考察。本論文為紅外光譜用于疾病篩查和中藥飲片所含指標(biāo)成分的同時(shí)測定等研究提供了方法學(xué)參考。研究內(nèi)容一:采用FTIR/ATR光譜建立不同的校準(zhǔn)模型,并應(yīng)用于新生兒的苯丙酮尿癥篩查。在本課題組過往研究的基礎(chǔ)上,通過串聯(lián)質(zhì)譜法測定69例干血斑樣品中苯丙氨酸(Phe)和酪氨酸(Tyr)的濃度,并使用FTIR/ATR采集樣品的光譜。通過平滑、求導(dǎo)、矢量歸一化、凹式Rubberband等方法對光譜進(jìn)行預(yù)處理,比較偏最小二乘法(PLS)、核偏最小二乘法(KPLS)、多模型共識偏最小二乘法(cPLS)及cKPLS四種模型構(gòu)建樣品中Phe濃度的校準(zhǔn)模型,以決定定系數(shù)(R2),均方根誤差(RMSE),平均相對誤差(MRE)等來評估所獲得的模型。結(jié)果發(fā)現(xiàn),經(jīng)過比較,cKPLS引入核方法和多模型共識,cKPLS模型表現(xiàn)最好,產(chǎn)生的結(jié)果更準(zhǔn)確和穩(wěn)定,為建立穩(wěn)健的FTIR/ATR光譜模型及解決FTIR/ATR光譜中其他復(fù)雜的校準(zhǔn)提供了新方法,并成功地應(yīng)用于預(yù)測新生兒干血片中Phe濃度,為苯丙酮尿尿癥篩查提供參考。研究內(nèi)容二:利用近紅外光譜法結(jié)合主成分分析和聚類分析對甘草進(jìn)行產(chǎn)地鑒別。對原始光譜進(jìn)行預(yù)處理后,分別用SPSS或MATLAB程序進(jìn)行主成分分析和聚類分析,對比兩種方法的聚類結(jié)果,發(fā)現(xiàn)SPSS聚類的效果優(yōu)于MATLAB程序�？梢园l(fā)現(xiàn),當(dāng)SPSS的類間距離為5.0時(shí),40份實(shí)驗(yàn)樣品可分為四大類:Ⅰ類為所有內(nèi)蒙古樣品;Ⅱ類為5份甘肅樣品、8份內(nèi)蒙古樣品、2份新疆,1份寧夏聚在一起;Ⅲ類為1份甘肅,1份新疆,1份內(nèi)蒙古;Ⅳ類為10份甘肅樣品同6份內(nèi)蒙古樣品、2份新疆、1份寧夏聚為一類;而用MATLAB程序做聚類分析時(shí),其類間距離為50時(shí),40份實(shí)驗(yàn)樣品被分成了八大類,而且沒有一個(gè)產(chǎn)地聚類的比較集中。研究內(nèi)容三:利用近紅外光譜和偏最小二乘法建立一種能夠同時(shí)快速測定甘草中甘草苷和甘草酸含量的定量校正模型。以HPLC所測樣品中甘草苷和甘草酸的含量為參考值,利用近紅外分析技術(shù),將參考值與樣品的近紅外光譜進(jìn)行關(guān)聯(lián),采用主成分(PCA)結(jié)合偏最小二乘法(PLS)建立甘草中甘草苷和甘草酸的定量分析模型。甘草苷的最優(yōu)模型結(jié)果為:校正決定系數(shù)(R2),驗(yàn)證集的決定系數(shù),校正均方差(RMSEC)和驗(yàn)證均方差(RMSEP)分別為0.9522,0.9305,0.0004和0.0017;甘草酸的最優(yōu)模型結(jié)果為:校正集的決定系數(shù),驗(yàn)證集的決定系數(shù)(R2)、校正均方差(RMSEC)和驗(yàn)證均方差(RMSEP)為0.9766,0.9591,0.0006和0.0021。結(jié)果發(fā)現(xiàn),利用近紅外光譜技術(shù)建立的定量分析模型,能夠?qū)Ω什葜懈什蒈蘸透什菟岬暮窟M(jìn)行快速無損的測定。
[Abstract]:In this paper, a qualitative and quantitative model based on Fourier Transform Infrared Spectroscopy-Attenuated Total Refraction (FTIR-ATR) and near infrared spectroscopy was developed for rapid screening of neonatal phenylketonuria (PKU) and simultaneous determination of glycyrrhizin and glycyrrhizin in licorice slices. The content of oxalic acid was determined by smoothing, derivative, principal component analysis, non-information variable elimination, normalization and other pretreatment methods. Partial least squares (PLS), consensus Partial Least Squares (cPLS), kernel partial least squares (Kernel Partial Least Squares) and multi-model consensus kernel bias were used. Quantitative calibration model of target component content was established by consensus Kernel Partial Least Squares, and the performance of the model was evaluated by various evaluation indexes. FTIR/ATR spectroscopy was used to establish different calibration models for neonatal phenylketonuria screening. Based on the previous studies of our group, the concentrations of phenylalanine (Phe) and tyrosine (Tyr) in 69 dry blood spot samples were determined by tandem mass spectrometry, and the spectra were collected by FTIR/ATR. Concave Rubberband and other methods were used to pre-process the spectra. The calibration models of Phe concentration in samples were constructed by comparing partial least squares (PLS), kernel partial least squares (KPLS), multi-model consensus partial least squares (cPLS) and cKPLS. The calibration models were used to determine the constant coefficients (R2), root mean square error (RMSE) and mean relative error (MRE). The results show that the cKPLS model performs best by introducing the nuclear method and multi-model consensus. The results are more accurate and stable. It provides a new method for establishing a robust FTIR/ATR spectral model and solving other complex calibration problems in FTIR/ATR spectroscopy. It has been successfully applied to predict the Phe concentration in the neonatal dry blood tablets and converting it into benzene. Content 2: Using near infrared spectroscopy combined with principal component analysis and clustering analysis to identify the origin of Glycyrrhiza uralensis. After pretreatment of the original spectrum, principal component analysis and clustering analysis were carried out by SPSS or MATLAB program, and the clustering results of the two methods were compared. It was found that the effect of SPSS clustering was better. In MATLAB program, it can be found that when the SPSS class spacing is 5.0, 40 experimental samples can be divided into four categories: I for all samples in Inner Mongolia; II for 5 samples in Gansu, 8 samples in Inner Mongolia, 2 samples in Xinjiang, 1 sample in Ningxia together; III for Gansu, 1 sample in Xinjiang, 1 sample in Inner Mongolia; IV for 10 samples in Gansu and 6 samples in Inner Mongolia. Two samples from Xinjiang and one sample from Ningxia were clustered into one group, while 40 samples were classified into eight groups when the distance between classes was 50 by MATLAB, and there was no clustering in any place of origin. Quantitative calibration model of glycyrrhizin content was established. The contents of glycyrrhizin and glycyrrhizin in the samples determined by HPLC were taken as reference values. The near infrared spectroscopy was used to correlate the reference values with the near infrared spectra of the samples. The quantitative analysis model of glycyrrhizin and glycyrrhizin in the samples was established by principal component analysis (PCA) combined with partial least squares (PLS). The optimal model results are: Corrected Decision Coefficient (R2), Verification Set Decision Coefficient (RMSEC), Corrected Mean Variance (RMSEC) and Verification Mean Variance (RMSEP) are 0.9522, 0.9305, 0.0004 and 0.0017 respectively; The optimal model results of glycyrrhizic acid are: Corrected Decision Coefficient (R2), Corrected Mean Variance (RMSEC) and Verification Mean Variance (RMSEP) are 0.9522, 0.9305, 0.0004 and 0.0017 respectively. 9766,0.9591,0.0006 and 0.0021. The results showed that the quantitative analysis model established by near infrared spectroscopy could be used to determine the content of glycyrrhizin and glycyrrhizic acid in licorice rapidly and nondestructively.
【學(xué)位授予單位】：廣東藥科大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：R284.1;O657.33
本文編號：2189797
[Abstract]:In this paper, a qualitative and quantitative model based on Fourier Transform Infrared Spectroscopy-Attenuated Total Refraction (FTIR-ATR) and near infrared spectroscopy was developed for rapid screening of neonatal phenylketonuria (PKU) and simultaneous determination of glycyrrhizin and glycyrrhizin in licorice slices. The content of oxalic acid was determined by smoothing, derivative, principal component analysis, non-information variable elimination, normalization and other pretreatment methods. Partial least squares (PLS), consensus Partial Least Squares (cPLS), kernel partial least squares (Kernel Partial Least Squares) and multi-model consensus kernel bias were used. Quantitative calibration model of target component content was established by consensus Kernel Partial Least Squares, and the performance of the model was evaluated by various evaluation indexes. FTIR/ATR spectroscopy was used to establish different calibration models for neonatal phenylketonuria screening. Based on the previous studies of our group, the concentrations of phenylalanine (Phe) and tyrosine (Tyr) in 69 dry blood spot samples were determined by tandem mass spectrometry, and the spectra were collected by FTIR/ATR. Concave Rubberband and other methods were used to pre-process the spectra. The calibration models of Phe concentration in samples were constructed by comparing partial least squares (PLS), kernel partial least squares (KPLS), multi-model consensus partial least squares (cPLS) and cKPLS. The calibration models were used to determine the constant coefficients (R2), root mean square error (RMSE) and mean relative error (MRE). The results show that the cKPLS model performs best by introducing the nuclear method and multi-model consensus. The results are more accurate and stable. It provides a new method for establishing a robust FTIR/ATR spectral model and solving other complex calibration problems in FTIR/ATR spectroscopy. It has been successfully applied to predict the Phe concentration in the neonatal dry blood tablets and converting it into benzene. Content 2: Using near infrared spectroscopy combined with principal component analysis and clustering analysis to identify the origin of Glycyrrhiza uralensis. After pretreatment of the original spectrum, principal component analysis and clustering analysis were carried out by SPSS or MATLAB program, and the clustering results of the two methods were compared. It was found that the effect of SPSS clustering was better. In MATLAB program, it can be found that when the SPSS class spacing is 5.0, 40 experimental samples can be divided into four categories: I for all samples in Inner Mongolia; II for 5 samples in Gansu, 8 samples in Inner Mongolia, 2 samples in Xinjiang, 1 sample in Ningxia together; III for Gansu, 1 sample in Xinjiang, 1 sample in Inner Mongolia; IV for 10 samples in Gansu and 6 samples in Inner Mongolia. Two samples from Xinjiang and one sample from Ningxia were clustered into one group, while 40 samples were classified into eight groups when the distance between classes was 50 by MATLAB, and there was no clustering in any place of origin. Quantitative calibration model of glycyrrhizin content was established. The contents of glycyrrhizin and glycyrrhizin in the samples determined by HPLC were taken as reference values. The near infrared spectroscopy was used to correlate the reference values with the near infrared spectra of the samples. The quantitative analysis model of glycyrrhizin and glycyrrhizin in the samples was established by principal component analysis (PCA) combined with partial least squares (PLS). The optimal model results are: Corrected Decision Coefficient (R2), Verification Set Decision Coefficient (RMSEC), Corrected Mean Variance (RMSEC) and Verification Mean Variance (RMSEP) are 0.9522, 0.9305, 0.0004 and 0.0017 respectively; The optimal model results of glycyrrhizic acid are: Corrected Decision Coefficient (R2), Corrected Mean Variance (RMSEC) and Verification Mean Variance (RMSEP) are 0.9522, 0.9305, 0.0004 and 0.0017 respectively. 9766,0.9591,0.0006 and 0.0021. The results showed that the quantitative analysis model established by near infrared spectroscopy could be used to determine the content of glycyrrhizin and glycyrrhizic acid in licorice rapidly and nondestructively.
【學(xué)位授予單位】：廣東藥科大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：R284.1;O657.33
【參考文獻(xiàn)】
相關(guān)期刊論文 前10條
1 吳敏;張偉濤;田沛榮;凌曉鋒;徐智;;正常人體甲狀腺體表傅里葉紅外光譜圖的特征分析[J];光譜學(xué)與光譜分析;2016年10期
2 侯湘梅;張磊;岳洪水;鞠愛春;葉正良;;基于近紅外光譜分析技術(shù)的丹參多酚酸大孔吸附樹脂柱色譜過程監(jiān)測方法[J];中國中藥雜志;2016年13期
3 李云;畢宇安;王振中;蕭偉;;近紅外光譜技術(shù)在熱毒寧注射液梔子提取液濃縮過程中的應(yīng)用[J];中國實(shí)驗(yàn)方劑學(xué)雜志;2016年12期
4 毛佩芝;楊凱;金葉;劉雪松;王龍虎;;近紅外光譜法快速測定野菊花藥材中水分及蒙花苷含量[J];中國現(xiàn)代應(yīng)用藥學(xué);2015年12期
5 楊佒雯;張錦水;朱秀芳;謝登峰;袁周米琪;;隨機(jī)森林在高光譜遙感數(shù)據(jù)中降維與分類的應(yīng)用[J];北京師范大學(xué)學(xué)報(bào)(自然科學(xué)版);2015年S1期
6 閆蔚;曾柏淋;王淑美;孟江;梁生旺;;6種硫酸鹽類礦物藥中紅外鑒別[J];中國實(shí)驗(yàn)方劑學(xué)雜志;2015年20期
7 吳紅梅;王祥培;楊燁;徐鋒;;液相色譜指紋圖譜技術(shù)在中藥鑒定學(xué)教學(xué)中的應(yīng)用探討[J];貴陽中醫(yī)學(xué)院學(xué)報(bào);2015年05期
8 馬丹;顧志榮;甘玉偉;趙克加;郭玫;;唐古特大黃及其不同炮制品的近紅外光譜分析[J];中藥材;2015年09期
9 周文婷;林萍;王海霞;姬生國;;巴戟天藥材中耐斯糖含量近紅外光譜測定方法的建立[J];井岡山大學(xué)學(xué)報(bào)(自然科學(xué)版);2015年05期
10 楊天鳴;張璐;付海燕;李鶴東;姜杜;周蓉;;不同產(chǎn)地甘草的近紅外指紋圖譜模式識別鑒別方法[J];亞太傳統(tǒng)醫(yī)藥;2015年14期
相關(guān)碩士學(xué)位論文 前1條
1 孔慶明;神經(jīng)網(wǎng)絡(luò)在食用油質(zhì)量近紅外光譜分析中的應(yīng)用研究[D];哈爾濱商業(yè)大學(xué);2012年
，本文編號：2189797
本文鏈接：http://sikaile.net/kejilunwen/huaxue/2189797.html
最近更新
教材專著
