Pure Ion Chromatograms Combined with Advanced Machine Learning Methods Improve Accuracy of Discriminant Models in LC-MS-Based Untargeted Metabolomics

Abstract

Untargeted metabolomics based on liquid chromatography coupled with mass spectrometry (LC-MS) can detect thousands of features in samples and produce highly complex datasets. The accurate extraction of meaningful features and the building of discriminant models are two crucial steps in the data analysis pipeline of untargeted metabolomics. In this study, pure ion chromatograms were extracted from a liquor dataset and left-sided colon cancer (LCC) dataset by K-means-clustering-based Pure Ion Chromatogram extraction method version 2.0 (KPIC2). Then, the nonlinear low-dimensional embedding by uniform manifold approximation and projection (UMAP) showed the separation of samples from different groups in reduced dimensions. The discriminant models were established by extreme gradient boosting (XGBoost) based on the features extracted by KPIC2. Results showed that features extracted by KPIC2 achieved 100% classification accuracy on the test sets of the liquor dataset and the LCC dataset, which demonstrated the rationality of the XGBoost model based on KPIC2 compared with the results of XCMS (92% and 96% for liquor and LCC datasets respectively). Finally, XGBoost can achieve better performance than the linear method and traditional nonlinear modeling methods on these datasets. UMAP and XGBoost are integrated into KPIC2 package to extend its performance in complex situations, which are not only able to effectively process nonlinear dataset but also can greatly improve the accuracy of data analysis in non-target metabolomics.

Keywords: KPIC2; LC–MS; Pure Ion Chromatogram; UMAP; XGBoost.

Scheme 1

Schematic diagram of the proposed data analysis pipeline for complex liquid chromatography-mass spectrometry (LC–MS)-based untargeted metabolomics. It can be divided into five parts: the extraction of metabolites from samples, LC–MS analysis, data preprocessing, visualization and statistical analysis. In the study, K-means-clustering-based Pure Ion Chromatogram extraction method version 2.0 (KPIC2) is used to extract pure ion chromatograms. The samples from different groups are visualized through uniform manifold approximation and projection (UMAP). Extreme gradient boosting (XGBoost) is used to build discriminant models and screen differential metabolites.

Figure 1

Venn diagram of the numbers of features in liquor dataset extracted by KPIC2 and XCMS. There are 259 features that are unique in KPIC2, and 231 features are also unique in XCMS. There are 504 features that are extracted by both KPIC2 and XCMS, which indicated the reliability of extraction results.

Figure 2

Venn diagram of the numbers of features in LCC dataset extracted by KPIC2 and XCMS. There are 433 features that are unique in KPIC2, and 287 features are also unique in XCMS. There are 958 common features detected by both KPIC2 and XCMS, which indicated the reliability of extraction results.

Figure 3

The receiver operating characteristic (ROC) curves of XGBoost models on the liquor dataset and the LCC dataset. Each color represents a class. (A) The ROC curve of XGBoost model trained by features of KPIC2 on the liquor dataset; (B) the ROC curve of XGBoost model trained by features of XCMS on the liquor dataset; (C) the ROC curve of XGBoost model trained by features of KPIC2 on the LCC dataset; (D) the ROC curve of XGBoost model trained by features of XCMS on the LCC dataset.

Figure 4

Visualization of the liquor dataset by PCA, t-SNE and UMAP. Each shape represents a sample. (A) The PCA plot based on features extracted by KPIC2 of the liquor dataset, and the percentage of variance explained by each selected component is displayed on the axis; (B) the t-SNE plot based on features extracted by KPIC2 of the liquor dataset; (C) the UMAP plot based on features extracted by KPIC2 of the liquor dataset.

Figure 5

Visualization of the LCC dataset by PCA, t-SNE and UMAP. Each shape represents a sample. (A) The PCA plot is based on features extracted by KPIC2 of the LCC dataset, and the percentage of variance explained by each selected component is displayed on the axis; (B) the t-SNE plot is based on features extracted by KPIC2 of the LCC dataset; (C) the UMAP plot is based on features extracted by KPIC2 of the LCC dataset.