Fig. 4

Predicted neoantigen load correlates with mutation load, while immune infiltration is highest in CMS1. a Among the 33 exome-sequenced tumors, there was a strong correlation between the number of exonic mutations (including synonymous) and predicted neoantigens. b There was not a one-to-one correspondence among the number of mutations (left), mutations predicted to create neoantigens (middle), and predicted neoantigens (right) per tumor (each tumor is indicated with a separate color), showing that individual mutations generate several neoantigens per tumor. c The individual mutations (represented by a dot) predicted to create most neoantigens per tumor (horizontal axis) were typically not recurrent in a large proportion of the tumors (vertical axis). However, several mutations (indicated by colors) created neoantigens in several tumors. d The number of amino acid changing mutations per tumor (horizontal axis) was not associated with the level of infiltration of cytotoxic lymphocytes (Spearman’s correlation −0.06, P = 0.8). e In contrast, among all MSI+ tumors in Norwegian series I, infiltration of cytotoxic lymphocytes and the ESTIMATE immune-score, as well as PD-1 and JAK-STAT signaling (based on gene expression) were significantly higher in CMS1 compared to CMS2-4 (Welch’s t-test; although high also in CMS4)