Hierarchical clustering of protein expression data. (A) Heatmap of normalized protein expression profiles. (B) Hierarchical clustering of non-normalized protein expression using averages clustering on a Euclidean distance matrix. Clusters were tested for statistical significance using multiscale bootstrapping, which revealed 4 statistically significant clusters (indicated by red boxes): 1 for each neutrophil subset and 1 for the banded and segmented neutrophils together. P values are approximately unbiased (AU; ie, an AU P value of .95 indicates that the existence of 2 clusters is supported by the data with 95% confidence). Numbers between brackets indicate the standard error of the AU P value estimation.