Separation of treatment groups by unsupervised principal component analysis (PCA) and hierarchical clustering analysis (HCA).

<p>(A) PCA score plot showing a three dimensional visualization of similarities and differences between each of the samples. Each data point corresponds to one sample with a 3-D (the first three principal components) projection of 1,240 metabolites. Clustering and separation of all samples indicates a gene dose trajectory in metabolite profile across groups. KO and WTA separated from WT along both PC1 (X-axis), which accounts for 39.4% of total variance. (B) PCA-loading plot showing 1,240 weighed metabolites contributing to the separation along PC1 and PC2 seen in score plot in <i>panel A</i>. Metabolites are colored by mass and shaped by compound. Identified metabolites contributing most strongly to the separation are indicated. Xanthine, hypoxanthine, xanthosine and urea were identified as up-regulated metabolites contributing dominantly to the separation along PC1, while uric acid and allantoin are among the down-regulated features that contribute most significantly to the separation along PC1. (C) Unsupervised HCA, based on the Pearson correlation, shows clustered within-group expression patterns for 1,240 quantified plasma features and significant differences in between-group expression. Feature intensity is depicted as a heat map, ranging from cold (blue) to hot (red) and was visualized using GeneSpring MS 1.2 software. Notably, mouse groups co-cluster with no-or-low XOR activity (<i>xor</i> KO and allopurinol-treated <i>xor</i> WT) and are clearly distinguished from mice with relatively higher XOR activity (<i>xor</i> HET and WT).