主成分分析(PCA)和偏最小二乘法(PLS)是对变量数超过样本数量或变量之间存在多重共线性的组学数据进行可视化、回归、分类和特征选择的常用方法。

PLS和正交偏最小二乘法(OPLS)是有监督的模式，它们使用偏最小二乘回归建立代谢物表达量与样本类别之间的关系模型，实现对样品类别的预测，是一种建模类型的方法 相较而言，OPLS能够分别对相关因子和不相关变异进行建模，虽然计算方式与PLS相同，但OPLS具有更强的解释性。

而且，当无监督(PCA)无法很好地区分组间样本时，PLS-DA可以实现有效分离。并且PLS-DA和OPLS-DA所构建的分类预测模型，可进一步用于识别更多的样本类别，这是探索性的PCA方法无法做到的。

另外，PLS-DA和OPLS-DA所构建的分类模型中的载荷图可用于衡量各代谢物组分对样本分类判别的影响强度和解释能力，辅助标志代谢物的筛选。 ## 实例解读

OPLS-DA loading plots for different mulberry cultivars

The metabolites with loadings that were distant from the origin on the OPLS-DA-loading plots were inferred to make the greatest contribution to class separation. Organic acids, lipids, and flavonols were major contributors to the separation among the three mulberry fruits.

该数据是通过液相色谱-高分辨率质谱(LC-HRMS)来研究年龄、体重指数(bmi)和性别对尿液中代谢物浓度的影响，是一个list具体包含：

Urine samples were analyzed by using an LTQ Orbitrap in the negative ionization mode. A total of 109 metabolites were identified or annotated at the MSI level 1 or 2. After retention time alignment with XCMS, peaks were integrated with Quan Browser. Signal drift and batch effect were corrected, and each urine profile was normalized to the osmolality of the sample. Finally, the data were log10 transformed. - Thevenot(2015)3

由于目的是识别给定数据集的特征，而不是建模预测未知数据的分类，在这里将所有数据均作为训练集构建模型。