Association between HLA-based metrics and survival outcomes of 1,443 cancer patients [16].

(A) and (B) summarize the results using between-individual HLA-I distance metrics. (A)-log10(p-value) of kernel regression using five different distance matrices with or without covariates. The covariates include age, log-transformed mutation burden, and drug type (CTLA-4, PD-1, or both). The grey horizontal line corresponds to p-value 0.05. (B) The dendrogram of hierarchical clustering where the distance is defined by “dist_DePTH_breadth”, and the survival curves of the two subgroups identified by cutting the dendrogram tree. The p-value is computed from Cox proportional hazard regression. (C) -log10(p-value) from Cox regression for the associations between individual-level HLA-I heterozygosity metrics and survival outcome, with or without covariates. The covariates include age, log-transformed mutation burden, and drug type (CTLA-4, PD-1, or both). The grey horizontal line corresponds to p-value 0.05.