本课题欲解决中药代谢组学研究中如何应用药学统计学高效、准确的获得生物标记物的问题。应用芫花、草乌和雷公藤三种中药肾毒性进行代谢组学研究得到的图谱信息以及收集已有的代谢组学数据，建立图谱数据库。根据中药代谢组学数据自身特点，对一个新的统计估算方法即Local False Discovery Rate(LFDR)应用于中药代谢组学生物标记物识别的问题进行研究，并应用此算法和常用的模式识别统计方法（如主成分分析、偏最小二乘判别分析、非线型映射、聚类分析等）建立中药代谢组学生物标记物识别程序包。应用LFDR估计方法作为生物标记物识别的量度，不仅可以找到生物标记物，而且还可以量化选择它的正确性。 与此同时，本课题对改进LFDR估计算法进行深入研究，降低LFDR做出判断的错误率以便更有效的说明每一个生物标记物。对于LFDR估计算法的深入研究和应用，为中药代谢组学生物标记物识别问题提供了一个新的思路。

In this project, we will focus how to obtain biomarker effectively and precisely by using pharmaceutical statistical models. Based on metabolic Spectrogram for Yuanhua, Kusnezoff Monkshood and Lei Gongteng, as well as the collected metabolic data for studying their renal toxicity, Spectrogram database will be built. Considering metabolic data characters and information of Traditional Medicine Chinese (TCM), we will build TCM metabolic biomarker identification （TCMBI） system not only by using a novel statistical method, Local false discovery rate (LFDR) estimation method but also by using some current pattern recognition methods, such as Principle component analysis (PCA),Partial least square-discriminant analysis(PLS-DA), Nonlinear model (NML), Clustering analysis(CA) and soft independent modeling of class analogies (SIMCA). Using LFDR as a measure in the problem of TCMBI, we can not only select biomarkers but also obtain the possibility that our selection is correct. Furthermore, we will deeply study LFDR and develop the LFDR estimation method to improve the precise of estimation to increase the accuracy of biomarker selection in the TCMBI system.The study of LFDR method will provide a new statistical idea in the area of TCM metabolism.