Machine Learning Algorithms Identify Pathogen-Specific Biomarkers of Clinical and Metabolomic Characteristics in Septic Patients with Bacterial Infections

Figure 2

Identification of clinical and metabolomic features associated with sepsis at hospital admission. (a) Performance of feature selection models based on variance threshold, MIC, and relief for the prediction of sepsis against all other controls. Image illustrated area under the receiver operating characteristic curve (AUC) values change depending on the number of features. (b) Sensitivity and specificity of the selected features by receiver operating characteristic analysis. (c) Pathway analysis of the metabolites filtered by the MIC model. (d) Enrichment analysis of blood disease-associated metabolites.

(a)

(b)

(c)

(d)