Fig. 5

Similarity of metabolic profiles from HFS and FD patients (HFS samples in red; FD samples in green; control samples in white). (a) Unsupervised PCA of all 5248 compounds that survived quality and quantity filters separates patient samples from controls samples, and this is largely based on principal component 1 (compare Fig. 3a. b Unsupervised hierarchical clustering confirms that patient samples are biochemically distinct from control samples, and additionally suggests that HFS patients and FD patients differ in their overall metabolomic profiles. (c) To-scale scheme visualizing all 5248 compounds (large square) in relation to the number of disease-specific compounds as specified. (d) Compound-specific fold changes (mean value for disease samples divided by mean value for control samples) for the 81 compounds which differed significantly from controls in both HFS and FD (D₁), in only HFS (D₂), or in only FD (D₃)