Statistical approach to metabonomic analysis of rat urine following surgical trauma
Statistical approach to metabonomic analysis of rat urine following surgical trauma
复制标题
DOI：
10.1002/cem.972
复制
发表时间：
2006-03-01
影响因子：
2.4
通讯作者：
Dey, Dipak K.
中科院分区：
文献类型：
--
作者：
Ghosh, Samiran;Dennis, W. H.;Dey, Dipak K.
急性创伤通常与人类多个器官系统的逐渐恶化有关，并且是创伤护理单位死亡的主要原因。先前的研究表明，多器官衰竭可能与不受控制的全身炎症有关。然而，其因果机制仍不清楚。当前评估创伤患者状态和预测结果的方法基于各种解剖学和/或生理学评分模型。虽然有用，但它们是劳动密集型的，并且不允许实时分析患者状态。在这项研究中，我们使用急性创伤的啮齿动物模型开发了一种基于代谢组学的方法，以确定对照大鼠和经历过手术创伤的大鼠的尿液代谢物的定量和定性特征之间是否存在统计学上的显着差异。这种方法结合了统计、分析和计算工具，以便识别创伤特有的代谢物，并可用于预测创伤结果。统计分析显示创伤组和对照尿代谢组之间存在显着差异。偏最小二乘 (PLS) 和主成分分析 (PCA) 结合逻辑回归 (LR) 用于将受试者分类为控制组或创伤组，准确率超过 80%。我们还表明，使用贝叶斯方法，我们可以对受创伤的受试者或具有给定可信区间的控制受试者进行分类。这些结果表明，代谢组学可能有助于量化和识别人类创伤状态以及创伤结果的生物标志物。版权所有 (c) 2007 John Wiley & Sons, Ltd.
Acute trauma is often associated with the progressive deterioration of multiple organ systems in humans and is the leading cause of death in trauma care units. Previous studies have suggested that multiple organ failure is likely related to uncontrolled systemic inflammation. However the causal mechanisms remain unknown. Current methods of assessing trauma patient status and predicting outcome are based on a variety of anatomical and/or physiological scoring models. While being useful, these are labor intensive and do not allow for real-time analysis of patient status. In this study, we have developed a metabonomic based approach using a rodent model of acute trauma in order to determine whether statistically significant differences exist between the quantitative and qualitative profile of urinary metabolites of control rats and rats that have experienced surgical trauma. This approach incorporates statistical, analytical, and computational tools in order to identify metabolites that are unique to trauma and maybe used to predict trauma outcome. Statistical analysis showed significant differences between the trauma and the control urinary metabonomes. Partial least square (PLS) and Principle component analysis (PCA) combined with Logistic regression (LR) were used to categorize subjects into either control or trauma with greater than 80% accuracy. We have also shown that using Bayesian methods that we could classify subjects being traumatic or control with a given credible interval. These results suggest that metabonomics may prove useful for quantifying and identifying biomarkers of trauma status as well as trauma outcome in humans. Copyright (c) 2007 John Wiley & Sons, Ltd.