气相色谱-质谱结合随机森林鉴定糖尿病小鼠经诺和龙治疗代谢轨迹

Identification of Metabolic Trajectory of Diabetic Mice in the Treatment of Repaglinide by Combining Gas Chromatography-Mass Spectrometry with Random Forest

采用气相色谱-质谱法(GC-MS)同时测定经诺和龙治疗的2型糖尿病KK-ay小鼠和C57BL/6J健康对照组小鼠尿液中的多种代谢物。利用随机森林算法对经诺和龙治疗后不同周期的2型糖尿病(T2DM)KK-ay小鼠的GC-MS数据进行统计分析,获得小鼠经治疗后的代谢轨迹图。样品预处理中,以核糖醇为内标,用甲醇除蛋白,尿酶除去尿素,经N2吹干后用肟化-硅烷化法进行衍生;GC-MS测定中,采用DB-5MS毛细管柱程序升温分离尿样中的多种成分,并结合NIST标准质谱库和标准品对尿液中代谢物进行定性定量分析,共鉴定出氨基酸、脂肪酸、有机酸、酯类和糖类等40种内源性代谢物质。将得到的数据输入随机森林进行建模分析,得到其治愈轨迹图。研究结果表明,诺和龙能有效地改善糖尿病小鼠的血糖代谢。

A simple, efficient and sensitive approach by combining gas chromatography-mass spectrometry（GC-MS） with random forest was developed for simultaneous determination of various endogenous metabolites in urine of two mice groups, i. e. , diabetes mice group and healthy control group. Ribitol was used as an internal standard. Methanol and urine enzyme were used for precipitating the proteins and urea, respectively. The oximation and silylation of samples dried by nitrogen were carried out before chromatographic analysis. Sufficient chromatographic separation for various metabolites was obtained with DB-5MS capillary column. Then, 40 endogenous metabolites including amino acids, fatty acids,organic acids, esters and carbohydrates were identified by NIST mass spectral library and standard substances. The data got from GC-MS profilings were further analyzed by random forest, which is a powerful tree ensemble machine learning statistic analysis tool. The results indicated that random forest is a useful approach to explore the relationships and differences hidden in the complex metabolomics data.