Fig. 3: Identification of hot and cold infiltration patterns.

a Stacked heat maps depicting an infiltration phenotype: the immune cell scores for a particular tissue, as estimated across the 3 deconvolution algorithms. Samples are in columns, with the ordering of samples along the x-axis identical for the 3 heatmaps. Blue, yellow, and red colors correspond to low to moderate to high estimated scores. Annotations assigning samples to hot, cold, and intermediate clusters are depicted along the top of each heatmap. In this example, cluster assignments were determined using consensus k-means clustering. Samples identified as hot or cold across each of the 3 deconvolution methods were identified as the “consensus” hot and cold groups, and considered for differential gene expression analysis. b Heatmaps display hot patterns across tissues and individuals. Clusters were determined by the consensus k-means clustering approach. Rows represent tissues and columns represent individuals. Red indicates the individual was labeled hot in that tissue type, while blue represents not hot (intermediate, cold, or missing data since an individual does not have a sample for every tissue type). Row and columns were clustered by Euclidean distance. Source data are provided as a Source Data file.