Skip to main content
. 2017 Aug 24;170(5):927–938.e20. doi: 10.1016/j.cell.2017.07.025

Figure S5.

Figure S5

Predicted Immunogenicity of HLA Class I Neoepitopes, Related to Figures 1 and 4 and Tables S1 and S4

(A) Predicted immunogenic properties of trunk (clonal) and private HLA-I neoepitopes. Positive immunogenicity scores have biochemical properties associated with higher immunogenicity that outweigh properties associated with lower immunogenicity, and vice versa for negative scores (Calis et al., 2013). Horizontal lines within violin plots show the median and interquartile range of the data distribution.

(B and C) Comparison between clonal and sub-clonal (including shared between two or more samples but not all) predicted immunogenicity of predicted binders and non-binders (two-sided Mann-Whitney rank test). Horizontal lines within violin plots show the median and interquartile range of the data distribution.

(D–F) Probability of an HLA-I neoepitope having immunogenic properties considering its clonality and HLA-I binding affinity using the neoepitope data in (A), (B), and (C), respectively. Clonal neoepitopes have a lower probability of having immunogenic properties than sub-clonal predicted binders (chi-square test, p = 0.02). For non-binders (NetMHC score > HLA-I specific cutoff), clonal neoepitopes have a higher probability of having immunogenic properties (chi-square test, p = 0.003), as well as peptides with higher HLA-I affinities (chi-square test, p = 0.0001), although the absolute differences are minor. No significant interaction between clonality and predicted HLA-I binding affinity was detected for either binders or non-binders. GLM = generalized linear model.