Fig. 6
Identification of TIME subtypes. A Distributions of leukocyte fractions (LFs) of human cancers. We analyzed LFs of 9692 tumors of 30 different cancer types, and 25th (LF = 0.086), 50th (LF = 0.171) and 75th (LF = 0.295) percentiles of LFs of these tumors are annotated in the figure. (B) Heatmap reveals two different clusters in the LF-high tumors. The clustering analysis was implemented using expression levels of 76 immune-related genes and infiltration levels of 22 types of immune cells. All values were scaled before clustering. Important immune cell types are labeled with arrows, and immune checkpoint and super categories of each gene are labeled on the left side. C Heatmap of scaled cluster centers of Cluster 1 and Cluster2. Values greater than 0 are in red, values less than 0 are in blue. Labels of 76 immune-related genes and 22 types of immune cells are annotated on the right side. Nearly all expression levels of the immune-related genes and immune cellular infiltrations are higher in Cluster1 than Cluster2. D TIME subtypes in LGG, COAD, CESC, KIRC, UVM and SKCM. Based on the LF levels and clustering results, we manually classified training samples of LGG, COAD, CESC, KIRC, UVM and SKCM into three different TIME subtypes, namely “Immune-Stimulation”, “Immune-Anergy” and “Immune-Exclusion” subtypes