Figure 2.

ElasticNETs metabolites relative importance. The heatmap reports the relative importance of the metabolites (columns) in each of the trained models (rows). Prior to visualization, metabolite coefficients were scaled per model by dividing them by the coefficients’ sum in each model to create the relative importance per model. Top: Metabolites were ordered based on the sum of their importance across all models. In addition, the models are clustered on the similarity between relative importance. Bottom: Categorized metabolic measures: “Amino acids”, “energy metabolism”, “inflammation”, “lipoproteins”, and “lipids and related measures”. Right: N. metabolites selected by the logistic ElasticNETs, Mean AUCs of the 5-FoldCV in a scale of purple.