China
Africa
Explore chapters and articles related to this topic
Targeted and Untargeted Metabolomics for Specific Food Intake Assessment
Published in Dale A. Schoeller, Margriet S. Westerterp-Plantenga, Advances in the Assessment of Dietary Intake, 2017
Carl Brunius, Huaxing Wu, Rikard Landberg
In untargeted LC–MS metabolomics experiments, multivariate data are generated. Typically, the number of variables (metabolic features) counts in the thousands to tens of thousands and thus far outweighs the number of samples. It is therefore often considered appropriate to analyze such data using multivariate modeling (Trygg et al. 2007). Among multivariate methods, a distinction is made between unsupervised and supervised methods. Unsupervised methods, such as principal component analysis (PCA) work by giving a representation of the measured, independent data (an X matrix; metabolomics data in this case) without the use of information from a dependent variable (normally a Y vector; such as dietary exposure in this case, although multiple responses are also possible). In supervised methods on the other hand, information from the Y variable(s) is used to guide the analysis of the X data, effectively resulting in a model relating them to each other (i.e., Y = f(X)). In the field of metabolomics, the partial least squares (PLS) family of methods, that is, regression and discriminant analysis using either PLS, OPLS, or sparse PLS, has become the de facto standard for supervised analysis (Trygg et al. 2007; Gromski et al. 2015). PLS and OPLS share the same analytical solutions and thus produce identical predictions, but in OPLS, the principal components are rotated to facilitate interpretation (Trygg et al. 2007). In sparse PLS modeling, on the other hand, a soft thresholding is employed to reduce the number of variables in the model, which often increases model performance (Lê Cao et al. 2008; Chung and Keles 2010). The overwhelming popularity of PLS methods is presumably related to the advantages they bring in terms of interpretation of the data. In fact, PCA and PLS share similarities in that they work by dimensionality reduction of the original data into sample scores, based on maintained variance in X for PCA and covariance between X and Y for PLS (Figure 18.1). The dimensionality-reduced scores can then easily be superimposed with variable loadings, and the scores and loadings together constitute the model components, where each component successively adds more explanatory power to the model. The combined biplot of scores and loadings from PCA and PLS analyses offers the possibility of simultaneous interpretation: The scores will reveal the structure or pattern in the data, whereas the loadings will give information as to which are the variables accounting for or driving the observed structure. In metabolomics, a combination of these strategies is often employed: Unsupervised PCA is performed initially to get a graphical overview of sample variability and identify potential outliers. Supervised PLS analysis is then performed to identify patterns in the metabolome corresponding to the research question at hand (Xia et al. 2012).
Effect of Wuzi Yanzong prescription on oligoasthenozoospermia rats based on UPLC-Q-TOF-MS metabolomics
Published in Pharmaceutical Biology, 2022
Zhimin Chen, Baohua Dong, Yunxiu Jiang, Ying Peng, Wenbing Li, Lingying Yu, Yongxiang Gao, Changjiang Hu
As shown in Figure 3, the PCA score plot was introduced to distinguish and evaluate the status of each group. In the positive and negative mode, the distribution of QC samples is concentrated, showing good reproducibility and reliability. The data of MG and other groups were well distinguished, and the distance between the CG and MG was far, indicating that the model was successfully established. The CG had some discrete data, which may be caused by individual differences and injection time differences of rats. OPLS-DA is a noise reduction method based on partial least squares method, which can remove some variations in the independent variables that were not related to the dependent variables, reduce the complexity of model, enhance the model forecast capability and better reveal the differences between groups (Zhou et al. 2016). The OPLS-DA model was used to further differentiate the differences among the groups. As shown in Figure 3, The OPLS-DA score plots (R2X (cum) = 0.469, R2Y (cum) = 0.970, Q2 (cum) = 0.592 in positive mode and R2X (cum) = 0.357, R2Y (cum) = 0.946, Q2 (cum) = 0.535 in negative mode) showed the prediction ability of the model is good, and the metabolites in each group were completely separated either on the positive or negative mode. The samples in the PG were more inclined to the CG, indicating that WYP had a callback effect on the abnormal endogenous metabolites.
Characterising phospholipids and free fatty acids in patients with schizophrenia: A case-control study
Published in The World Journal of Biological Psychiatry, 2021
Dongfang Wang, Xiaoyu Sun, Michel Maziade, Wei Mao, Chuanbo Zhang, Jingyu Wang, Bing Cao
To identify metabolites responsible for discrimination between schizophrenic patients and controls, orthogonal partial least-squares discriminant analysis (OPLS-DA) was performed using log10-transformed and auto scaled data to construct classification models (Bylesj et al. 2006). An internal seven-fold cross-validation was carried out to estimate the performance of the OPLS-DA model. The quality of the OPLS-DA models was described by two parameters (R2 and Q2). R2 represents the explanation capacity of the model, while Q2 stands for the predictive capacity of the model (Mahadevan et al. 2008). In addition to cross-validation, model validation was also performed using 300-iteration permutation tests. If the values of Q2 and R2 resulting from the original model were higher than the corresponding values from the permutation test, the model was considered valid (Mahadevan et al. 2008). OPLS-DA modelling was performed by the SIMCA-P software 14.1 (Umetrics, Umeå, Sweden). The variable importance in projection (VIP) values for each variable was obtained from the cross-validated model which represented the separation capacity for different groups. Moreover, the corresponding fold change (FC) was calculated to show the degree of variation in metabolite levels between groups.
Serum metabolomics of end-stage renal disease patients with depression: potential biomarkers for diagnosis
Published in Renal Failure, 2021
Dezhi Yuan, Tian Kuan, Hu Ling, Hongkai Wang, Liping Feng, Qiuye Zhao, Jinfang Li, Jianhua Ran
Partial Least Squares Discrimination Analysis (PLS-DA) or Orthogonal PLS-DA (OPLS-DA) is a supervised discriminant analysis statistical method. This method uses PLS-DA to establish a model of the relationship between the expression of metabolites and the sample category to realize the prediction of the sample category. Establish a PLS-DA model or OPLS-DA model for group comparisons, and calculate the variable importance for the projection (Variable Importance for the Projection, VIP) to measure the influence of the expression pattern of each metabolite on the classification of each group of samples. And interpretation capabilities, thereby assisting the screening of marker metabolites.