Figure 1

Hierarchical clustering identifies immune profiles within patient groups. Peripheral blood leukocyte populations were measured by flow cytometry. The number of cells/μl for each marker was determined directly or converted from TruCount tubes. All phenotype values were normalized against the mean of similarly measured and converted healthy volunteers (n = 40). Unsupervised clustering was performed using ten immune markers for each disease group (blue) and healthy volunteers (red; HV). The same HV cohort was used for all clustering analysis. Identification of major clusters is indicated at left. A row represents one subject and a column represents one of ten markers measured. The horizontal bar below each plot indicates immune markers decreased (blue) or increased (red) over the mean of the healthy volunteer cohort. (A) Clustering of patients with glioblastoma (GBM; n = 27). GBM patients were further identified based on the presences of pre-operative dexamethasone (DEX; purple), its absence (NONE; orange). HV indicated in green. (B) Clustering of patients with non-Hodgkin lymphoma (NHL; n = 28). (C) Clustering of renal cell carcinoma patients (RCC; n = 25). (D) Clustering of patients with acute lung injury with or at risk for sepsis (ALI; n = 23). ALI patients are further identified as those with (purple) or without (orange) confirmed sepsis as well as those that did (brown) or did not (pink) survive the episode.