Kasemsumran Sumaporn, Du Yi-Ping, Murayama Koichi, Huehne Martin, Ozaki Yukihiro
Department of Chemistry and Research Center for Near-Infrared Spectroscopy, School of Science and Technology, Kwansei Gakuin University, Sanda, Hyogo 669-1337, Japan.
Analyst. 2003 Dec;128(12):1471-7. doi: 10.1039/b307294k. Epub 2003 Nov 12.
Near-infrared (NIR) spectra in the 12,000-4,000 cm(-1) region were measured for phosphate buffer solutions containing human serum albumin (HSA), gamma-globulin, and glucose with various concentrations at 37 degrees C. Five levels of full factorial design were used to prepare a sample set consisting of 125 samples of three component mixtures. The concentration ranges of HSA, gamma-globulin and glucose were 0.00-6.00 g dl(-1), 0.00-4.00 g dl(-1) and 0.00-2.00 g dl(-1), respectively. The 125 sample data were split into two sets, the calibration set with 95 data and the prediction set with 30 data. The most informative spectral ranges of 4648-4323, 4647-4255 and 4912-4304 cm(-1) were selected by moving window partial least-squares regression (MWPLSR) for HSA, [gamma]-globulin, and glucose in the mixtures, respectively. For HSA, the correlation coefficient (R) of 0.9998 and the root mean square error of prediction (RMSEP) of 0.0289 g dl(-1) were obtained. For [gamma]-globulin, R of 0.9997 and RMSEP of 0.0252 g dl(-1) were obtained. The corresponding statistic values of glucose were 0.9997 and 0.0156 g dl(-1), respectively. These statistical values obtained by MWPLSR are highly significant and better than those calculated by using the regions reported in the literature. The results presented here show that MWPLSR can select the informative regions with a simple procedure and increase the power of NIR spectroscopy for simultaneous determination of the concentrations of HSA, [gamma]-globulin and glucose in the mixture systems.
在37℃下，对含有不同浓度人血清白蛋白（HSA）、γ-球蛋白和葡萄糖的磷酸盐缓冲溶液，测量了12,000 - 4,000 cm⁻¹区域的近红外（NIR）光谱。采用五水平全因子设计制备了由125个三组混合物样品组成的样本集。HSA、γ-球蛋白和葡萄糖的浓度范围分别为0.00 - 6.00 g dl⁻¹、0.00 - 4.00 g dl⁻¹和0.00 - 2.00 g dl⁻¹。将125个样本数据分为两组，校准集有95个数据，预测集有30个数据。通过移动窗口偏最小二乘回归（MWPLSR）分别为混合物中的HSA、γ-球蛋白和葡萄糖选择了最具信息性的光谱范围4648 - 4323、4647 - 4255和4912 - 4304 cm⁻¹。对于HSA，获得的相关系数（R）为0.9998，预测均方根误差（RMSEP）为0.0289 g dl⁻¹。对于γ-球蛋白，R为0.9997，RMSEP为0.0252 g dl⁻¹。葡萄糖的相应统计值分别为0.9997和0.0156 g dl⁻¹。通过MWPLSR获得的这些统计值非常显著，并且优于使用文献中报道的区域计算得到的统计值。此处给出的结果表明，MWPLSR可以通过简单的程序选择信息区域，并提高近红外光谱法同时测定混合体系中HSA、γ-球蛋白和葡萄糖浓度的能力。
