Figure 5.

Machine learning (ML) classification of group lasso-selected metabolites. Heatmap representation of metabolite importance score from the different ML models. Scores are scaled from 0 (not important) to 1 (very important) to the model’s performance. Asterisks denote compounds that have not been confirmed against a standard, but whose identity the analytical platform is confident in. ML operating characteristics are provided in online supplemental figure S11. GBM, generalised boosted models; LDA, linear discriminant analysis; PAM, prediction analysis for microarrays; RF, random forest; RLR, regularised logistic regression with elastic net regularisation; RPART, recursive partitioning and regression trees; SVM, support vector machine.