Research Paper Volume 13, Issue 10 pp 14322—14341

Aging-related markers in rat urine revealed by dynamic metabolic profiling using machine learning

Two algorithms based on machine learning were used for feature selection in the negative mode. (A) Least Absolute Shrinkage and Selector Operation (LASSO) algorithm in the training group. (B) Support Vector Machine-Recursive Feature Elimination (SVM-RFE) algorithm in the training group. (C) Venn diagrams demonstrated the overlap of metabolites in combination with two algorithms. (D) 14 candidate metabolites detected in the training group were validated in the test group using receiver operator characteristic (ROC) curve analysis. (E) Cluster analysis of 25 metabolites in both the negative and positive mode simultaneously selected from the LASSO and SVM-RFE algorithms in the training group.

Figure 3. Two algorithms based on machine learning were used for feature selection in the negative mode. (A) Least Absolute Shrinkage and Selector Operation (LASSO) algorithm in the training group. (B) Support Vector Machine-Recursive Feature Elimination (SVM-RFE) algorithm in the training group. (C) Venn diagrams demonstrated the overlap of metabolites in combination with two algorithms. (D) 14 candidate metabolites detected in the training group were validated in the test group using receiver operator characteristic (ROC) curve analysis. (E) Cluster analysis of 25 metabolites in both the negative and positive mode simultaneously selected from the LASSO and SVM-RFE algorithms in the training group.