Fig. 2

Partial least squares discriminant analysis (PLS-DA) for the 2–17-year-old cohorts based on the preprocessed original data. a Metabolomic profiling data. b Inflammatory protein mediator profiling data. c Combined biomarker profiling data. The three-dimensional PLS-DA score scatterplots show the distribution of observations (red dots, pediatric intensive care unit [PICU] sepsis patients; blue dots, emergency department [ED] sepsis patients; green dots, ED controls) in the three-dimensional space formed by PLS components (PLS1, PLS2, and PLS3). During the model construction, a discriminant plane (PLS component) was found in which the projected observations were well separated according to the class (PICU sepsis, ED sepsis, ED controls). The sphere describes the 95 % confidence interval of the Hotelling’s T ² distribution