Fig. 4: Positive selection of HLA-I genes.

a, Pan-cancer dN:dS ratios of HLA-I genes in the metastatic (left) and primary (right) dataset. The vertical lines represent the 5% and 95% confidence intervals (CIs) after ten randomizations, dots the maximum likelihood estimates and n the number of samples. b, Metastatic pan-cancer and gene-specific dN:dS ratios of HLA-I genes. The vertical lines represent the 5% and 95% CIs after ten randomizations and the dots the maximum likelihood estimates. c, Similar to b for the primary dataset. d,e, Representation of gene and cancer type-specific positive selection of HLA-I in the metastatic (d) and primary cohorts (e). f, Needle plots representing the pan-cancer distribution of somatic mutations along the HLA-A, HLA-B and HLA-C protein sequences in the metastatic dataset. Mutations are colored according to the inferred consequence type. Rectangles represent the Pfam³⁶ domains. g, Distribution of nonfocal LOH events along the autosomes in PANET tumors of the metastatic cohort (ticks on the x axis represent the chromosomal starting position). h, Distribution of focal LOH events along the autosomes in cervical cancer tumors of the metastatic cohort (ticks on the x axis represent the chromosomal starting position). i, Distribution of highly focal LOH events surrounding the HLA-I locus, spanning chromosome (chr) 6 from 25.1-Mb to 49.9-Mb genomic locations in the metastatic colorectal cancer cohort. Each bin represents 100 kb. Dashed horizontal lines represent the expected mean after randomization and vertical dashed lines highlight the HLA-I genomic locations. CDS pos., coding sequence position.