Figure 4

Comparison of urine metabolites by patient and risk group. Metabolites were K-means clustered followed by ascendant hierarchical clustering based on Euclidian distances with twenty-one metabolites excluded (0.25 < std dev). Metabolite clusters were also represented via a dendrogram displayed vertically for metabolites and another horizontally for patients. The data values of the permuted matrix were replaced by corresponding color intensities based on interquartile range with color scale of red to green through black resulting in a heat map. Patient identifiers and risk categories were replaced by color bars. Color bars on the top of the graph denote patient groups and bottom risk class.