Clustering of clinical parameters and HLA-DR and CD45RO surface staining. (A) Unsupervised clustering of the results from surface staining of CD4+ and CD8+ cells for the T-cell activation marker HLA-DR and the T-cell memory marker CD45RO. (B) Supervised clustering of immune response parameters according to response to therapy. Each column represents results for 1 patient. The first 5 rows describe clinical features of patients, including response to therapy (0 = No response; 1 = Response), site (U = University of California, Los Angeles; M = MD Anderson), sex (M = Male; F = Female), stage (3 = IIIc; 4 = IV), age, and metastasis (0 = Lung; 1 = Nodes; 2 = Liver; 3 = Subcutaneous; 4 = Skin; 5 = Stomach; 6 = Adrenal; 7 = Kidney). The rest of the rows represent test results at different time points after dosing. White represents no change (0 value) or absence of data (which is not included in the clustering algorithm). Red and shades of red represent positive changes on a scale of 1 to 4. Blue and shades of blue represent negative changes on a scale of 1 to 4. Data are normalized to a -4 to +4 scale as described in Table 2. All test results were included in the analysis, and the graph represents the results that best discriminate clinical responders (1108, 1111, 1113) from nonresponders, setting a 1.2-fold difference as the cutoff. Each column and row is presented as described in Figure 4, and the data are normalized to a -4 to +4 scale as described in Table 2. HLA = Human leukocyte antigen; EOS = End of study; QPCR = Quantitative polymerase chain reaction.