Analytical Sciences, Amgen Inc., Longmont, CO 80503, USA.
Anal Biochem. 2013 Mar 1;434(1):153-65. doi: 10.1016/j.ab.2012.11.018. Epub 2012 Dec 3.
Optical and vibrational spectroscopic techniques are important tools for evaluating secondary and tertiary structures of proteins. These spectroscopic techniques are routinely applied in biopharmaceutical development to elucidate structural characteristics of protein products, to evaluate the impact of processing and storage conditions on product quality, and to assess comparability of a protein product before and after manufacturing changes. Conventionally, the degree of similarity between two spectra has been determined visually. In addition to requiring a significant amount of analyst training and experience, visual inspection of spectra is inherently subjective, and any determination of comparability based on visual analysis of spectra is therefore arbitrary. Here, we discuss a general methodology for evaluating the suitability of numerical methods to calculate spectral similarity, and then we apply the methodology to compare four quantitative spectral similarity methods: the correlation coefficient, area of spectral overlap, derivative correlation algorithm, and spectral difference methods. While the most effective spectral similarity method may depend on the particular application, all four approaches are superior to visual evaluation, and each is suitable for assessing the degree of similarity between spectra.
光学和振动光谱技术是评估蛋白质二级和三级结构的重要工具。这些光谱技术在生物制药开发中被常规应用，以阐明蛋白质产品的结构特征，评估处理和储存条件对产品质量的影响，并评估制造变更前后蛋白质产品的可比性。传统上，通过目视确定两个光谱之间的相似程度。除了需要大量分析师的培训和经验外，光谱的目视检查本质上是主观的，因此基于光谱目视分析的任何可比性确定都是任意的。在这里，我们讨论了一种评估数值方法计算光谱相似性适用性的一般方法，然后我们将该方法应用于比较四种定量光谱相似性方法：相关系数、光谱重叠面积、导数相关算法和光谱差方法。虽然最有效的光谱相似性方法可能取决于特定的应用，但所有四种方法都优于目视评估，并且每种方法都适用于评估光谱之间的相似程度。
