Figure 4.

Putative network distribution of metabolites identified by different methods used to analyze cohort-based metabolomics data. The number of metabolites found in association with age (continuous outcome) and sex (binary outcome) from experimentally derived metabolomics studies (see text) for the different statistical methods applied was greater for traditional than for statistical learning models. Notably, the former identified metabolites that tended to be highly correlated with each other (Spearman rho ≥ 0.65), whereas the latter identified a more parsimonious number of metabolites distributed across the putative network of all highly intercorrelated metabolites. BON, Bonferroni; FDR, false discovery rate; LASSO, least absolute shrinkage and selection operator; SPLS, sparse partial least squares.