甾体激素的紫外光谱褶合变换可视化指纹谱鉴别研究

Study on convolution transform and visualization of ultraviolet spectroscopic information of steroid hormones

目的:建立一种基于紫外光谱法的甾体激素鉴别新方法。方法:运用褶合变换挖掘紫外光谱的特征信息并通过科学计算可视化技术获取相应指纹谱的表征,并利用夹角余弦法计算指纹谱之间的相似系数,考察方法的效能指标并对地塞米松等23种结构相近甾体激素进行鉴别。结果:本方法的精密度、耐用性均较好。不同类的甾体激素之间相似系数有较大差异,同类激素的烯键不同对相似系数的影响较大,不同侧链或者酯键对相似系数的影响较小。结论:褶合变换与科学计算可视化技术相结合可以拓展紫外光谱法在药物定性鉴别中的应用。

Objective： To establish a new analytical method for the identification of steroid hormones. Methods： Convolution transform combined with information visualization techniques were employed in the data mining of ultraviolet spectrum, resulting in the fingerprint spectrum of certain compounds. The similarity coefficient of fingerprint spectrum based on cosine algorithm was further used in the identification of 23 steroid hormones, ie. prednisolone, betamethasone, pregnendione, etc., after the method was validated. Results： The method showed acceptable precision and ruggedness. There were large differences between the similarity coefficients of different types of steroid hormones. The unsaturated bond contributed greatly to the differences between the similarity coefficients of same types of steroid hormones, while the side chain or its esterifieation had relatively small effect on the similarity coefficients. Conclusion： Convolution transform combined with information visualization techniques broadens the application of ultraviolet spectrum in the pharmaceutical identification.