透過您的圖書館登入
IP:3.215.77.96
  • 學位論文

壹維質子核磁共振代謝體圖譜之基線校正、圖譜還原與在電焊暴露代謝體特徵上的應用

Baseline Correction, Spectral Deconvolution of Metabolomic 1D 1H-NMR Spectra and Their Applications in Metabolomic Characterization of Welding Fumes Exposure

指導教授 : 曾宇鳳

摘要


代謝體研究的目標在於偵測並且定量細胞、組織和生物體內所有的內源性與外源性小分子代謝物。測量這一大群小分子代謝物的標型變化,可以讓我們了解到外界的刺激、環境的變化、基因的修飾或是藥物、疾病、營養等種種因素所造成的代謝機制改變以及其生理病理機轉。藉由高解析度的一維質子核磁共振圖譜,我們可以取得這些代謝物大量且高品質的資訊。因此,一維質子核磁共振實驗被廣泛的應用於代謝體的研究之中。 然而一維質子核磁共振圖譜基線的彎曲,會干擾各種代謝物成分的定量。針對複雜的一維質子核磁共振代謝體圖譜,目前所使用的基線校正方法極度仰賴有經驗的使用者,依照手動的方式去調整參數,以求達到最佳校正的結果。因此使用這些校正的方法,很容易會因為人為的因素而產生不同的校正結果,並造成代謝物定量上的偏差。我們為求達到自動化的需求並且客觀的校正一維質子核磁共振代謝體圖譜的基線,所以提出了一個新的基線校正方法,命名為BaselineCorrector。BaselineCorrector使用一個移動的視窗,自圖譜開頭依序滑移至圖譜尾端,計算並收集視窗中圖譜訊號強度的標準差。BaselineCorrector可以將圖譜每一個視窗中雜訊標準差的分布,模擬成一個衍生的卡方分布。藉著這個分布的模型,BaselineCorrector可以決定最適當的參數,將圖譜中的每一個點,以最少錯誤的方法分類成訊號點或雜訊點。由於不同圖譜間的雜訊都具有共通的分佈特性,因此BaselineCorrector可以正確識別不同種類圖譜的基線。除了常用的1D NOESY以及CPMG等一維質子核磁共振代謝體圖譜,BaselineCorrector也提供了額外的方法以校正diffusion-edited等核磁共振代謝體圖譜。藉著植基於分布模型的分類法,BaselineCorrector可以在基線校正過程中,成功的保留核磁共振代謝體圖譜中低強度的訊號,並且正確的處理複雜代謝體圖譜中所特有的寬廣之重疊訊號。 除了基線的問題之外,代謝體核磁共振圖譜中,各分子的訊號常有嚴重重疊的情況。分子訊號的重疊以及其位置的不確定性,會造成識別與定量個別代謝物組成上很大的問題。在解析這些重疊訊號時,曲線配適法比起合併處理法,更能保留圖譜的解析度。然而代謝體核磁共振圖譜中有無數的重疊訊號,如果使用手工進行曲線配適,將是非常耗時且困難的工作。目前使用的自動曲線配適法,在分析代謝體圖譜的重疊訊號時,經常必須侷限於圖譜特定的區域或有限的標的,才能確保其正確性。為了使用大量的標準品參考圖譜去完整解析代謝體圖譜中的重疊訊號,我們提出了一個雙重限制的矩陣解析法(double-bounded matrix factorization algorithm , DBMF)。我們的方法首先將單一標準品參考圖譜的訊號對齊到其附近被解析圖譜中的最強訊號,然後進行最小方差運算。在進行最小方差運算時,為了避免負的濃度值以及偽陽性的結果的出現,DBMF對解出的係數設了一個非負值限制以及一個上限。藉由這個雙重的限制,DBMF可以有效率的處理大量的標準品參考圖譜,並減少伴隨的共線性問題。我們使用混合溶液、細胞萃取液以及合成的圖譜,並使用高達535個單一標準品參考圖譜對DBMF作測試。結果發現不管是對簡單溶液或是對複雜混合物圖譜作解析時,在使用大量參考圖譜的情況下,DBMF都可以大幅增加定量的準確度,並減少偽陽性的結果。 基於代謝體研究在毒物學分析上的優越性,我們也使用代謝體學的方法針對電焊氣體暴露的毒性設計一個實驗。由於評估電焊氣體暴露的研究大幅受到電焊氣體複雜的組成、多重的分子標的、多變化的細胞作用、以及電焊工人的不同生活形態等因素所干擾,代謝體的研究可以綜合這些影響因素。本研究使用氫原子核磁共振光譜以及標型辨認的方法,對台灣造船廠的35個男性電焊工人及16個男性辦公人員的尿液樣本,測定其代謝體特徵。本研究也收集了這51個受試者的血液樣本,測定細胞激素和發炎標記的表現量。我們依電焊氣體暴露量的高低,檢驗電焊工人與辦公人員兩組間,其代謝物小分子和發炎標記的差異。電焊工地氣體顆粒測定顯示電焊工人暴露於濃度不等的鉻、鎳、錳等重金屬顆粒中。尿液代謝體組成的多變量統計分析顯示,電焊工人有較高濃度的甘胺酸、牛磺酸、三甲基甘胺酸�氧化三甲胺、絲胺酸、硫胱胺酸、馬尿酸鹽、葡萄糖酸鹽、肌酐酸以及丙酮,但卻有較低濃度的肌酸。在細胞激素和發炎標記中,則只有TNF-α在兩組間存有顯著差異。在有差異的代謝物中,在電焊工人較高濃度的甘胺酸、牛磺酸、三甲基甘胺酸,可能具有調控發炎以及氧化傷害的功能。代謝體的研究結果顯示,和電銲氣體暴露相關的代謝體變化,會受到抽菸的影響,但較不受到喝酒的影響。這個結果和之前抽菸會影響發炎標記的研究結果相符合,也和之前非代謝體分析的研究結果相呼應。這研究顯示代謝體特徵是一個有效的方法,可以顯示電焊氣體暴露和其他干擾因子總合的影響。

並列摘要


Metabolomic investigations are aimed to detect and quantitatively measure all of the endogenous and exogenous low molecular weight metabolites in cells, tissues, or organisms. Patterns of these large quantities of metabolites are able to provide considerable insight into dis-regulated metabolic pathways and underlying pathophysiological changes induced by external stimuli, environmental stress, disease processes, drugs, nutrition or genetic modification. High-resolution metabolomic 1D proton nuclear magnetic resonance (1H-NMR) spectra of biofluids provide rich and high-quality information about their metabolite compositions. Metabolomic studies based on 1H-NMR experiments are widely employed in biomedical research. However, baseline distortion in 1D 1H-NMR data complicates the quantification of individual components of biofluids in metabolomic experiments. Current 1D 1H-NMR baseline correction methods usually require manual parameter and filter tuning by experienced users to obtain desirable results from complex metabolomic spectra – thus becoming prone to correction variation and biased quantification. We present a novel alternative method, BaselineCorrector, for automatically estimating the baselines of 1D 1H-NMR metabolomic data. By collecting the standard deviations of spectral intensities, using a moving window to slide through a spectrum, BaselineCorrector can model the distribution of noise standard deviation as a derived chi-squared distribution in each window and then determine optimal parameters for least-error classification of signal and noise. Due to the universal property of noise distributions, BaselineCorrector can robustly recognize the baseline segments in various spectra. Using its classification model, BaselineCorrector is able to preserve low signal peaks and correctly handle wide, overlapping peaks in complex metabolomic spectra. In addition to the baseline problems, the overlapping signals of the 1D 1H-NMR metabolomic spectra and their position uncertainty cause a fundamental problem to identify/quantify the individual metabolite components. Curve fitting methods, compared to the binning methods, have the advantages of preserving spectral resolution to identifying/quantifying the overlapping signals. However, due to the numerous overlapping signals, manual curve fitting for these spectra is very time-consuming and challenging. The successful applications of current automatic curve fitting algorithms on metabolomic spectra are usually restricted to selected spectral segments or limited targets. To handle a large number of reference spectra and recover the complete signals of metabolomic NMR spectra, we proposed a double-bounded matrix factorization algorithm (DBMF). Our algorithm first applies a peak alignment method to assign reference signals to proper chemical shift positions. During the least-squares processes, DBMF set a non-negative lower bound to each deconvoluted concentration coefficient to prevent the negative predicted value and set an upper bound to reduce the false positive predictions. By the double-bounded method, DBMF can effectively manage a large reference set and reduce the associated problems of multicollinearity. We tested DBMF with the spectra of solution mixtures, cell extracts and synthetic data utilizing reference sets containing up to 535 spectra. DBMF is able to markedly increase the quantification accuracy and decrease the false positive predictions in both simple and complex spectra when large reference sets are used. Due to the advantages of metabolomic analysis on toxicology, we applied a metabolomic study to investigate the toxic effects of welding fumes exposure. The effect of air pollutants produced by welding processes on the health of welders is a major concern of occupational medicine. However, the complex composition of welding fumes, multiplicity of molecular targets, diverse cellular effects, and lifestyles associated with laborers vastly complicate assessment of welding fume exposure. The urinary metabolomic profiles of 35 male welders and 16 male office workers at a Taiwanese shipyard were characterized via 1H-NMR spectroscopy and pattern recognition methods. Blood samples for the same 51 individuals were also collected and the expression levels of the cytokines and other inflammatory markers were examined. This study dichotomized the welding exposure variable into high (welders) versus low (office workers) exposures to examine the differences of continuous outcome markers - metabolites and inflammatory markers - between the two groups. Fume particle assessments showed welders were exposed to different concentrations of chromium, nickel and manganese particles. Multivariate statistical analysis of urinary metabolomic patterns showed higher levels of glycine, taurine, betaine/TMAO, serine, S-sulfocysteine, hippurate, gluconate, creatinine, and acetone and lower levels of creatine among welders, while only TNF-α was significantly associated with welding fume exposure among all cytokines and other inflammatory markers measured. Of the identified metabolites, the higher levels of glycine, taurine, and betaine among welders were suspected to play some roles in modulating inflammatory and oxidative tissue injury processes. In this metabolomics experiment, it is discovered that the association of the identified metabolites with welding exposure was confounded by smoking, but not with drinking, which is a finding consistent with known modified response of inflammatory markers among smokers. The results correspond with prior studies that utilized non-metabolomic analytical techniques, and suggest that the metabolomic profiling is an efficient method to characterize the overall effect of welding fume exposure and other confounders.

參考文獻


(1) Clayton, T. A.; Lindon, J. C.; Cloarec, O.; Antti, H.; Charuel, C.; Hanton, G.; Provost, J. P.; Le Net, J. L.; Baker, D.; Walley, R. J.; Everett, J. R.; Nicholson, J. K. Nature 2006, 440 (7087), 1073-1077.
(2) Robertson, D. G. Toxicol Sci 2005, 85 (2), 809-822.
(3) Weljie, A. M.; Newton, J.; Mercier, P.; Carlson, E.; Slupsky, C. M. Anal Chem 2006, 78 (13), 4430-4442.
(4) Lindon, J. C.; Holmes, E.; Nicholson, J. K. Expert Rev Mol Diagn 2004, 4 (2), 189-199.
(5) Beckonert, O.; Keun, H. C.; Ebbels, T. M.; Bundy, J.; Holmes, E.; Lindon, J. C.; Nicholson, J. K. Nat Protoc 2007, 2 (11), 2692-2703.

延伸閱讀