Population-level distribution and putative immunogenicity of cancer neoepitopes

Abstract

Background: Tumor neoantigens are a driver of cancer immunotherapy response; however, current neoantigen prediction tools produce many candidates that require further prioritization for research/clinical applications. Additional filtration criteria and population-level understanding may help to produce refined lists of putative neoantigens. Herein, we show neoepitope immunogenicity is likely related to measures of peptide novelty and report population-level behavior of these and other metrics. Methods: We propose four peptide novelty metrics to refine predicted neoantigenicity: binding difference, paired similarity, homolog similarity, and microorganism similarity. We apply these metrics to tumor neoepitopes predicted from somatic missense mutations in The Cancer Genome Atlas (TCGA) and a cohort of melanoma patients, as well as to a small validation cohort of peptides with neoepitope-specific immune response data using an extension of pVAC-Seq [1]. Results: We show neoepitope burden varies across TCGA disease sites and HLA alleles, with surprisingly low repetition of neoepitope sequences across patients or neoepitope preferences among sets of HLA alleles. Only 19.9% of predicted neoepitopes across TCGA patients displayed novel binding change based on our binding difference criteria. Similarity of amino acid sequence was typically high between paired tumor-normal epitopes, but in 24.6% of cases, neoepitopes were more similar to other human peptides, or even to bacterial (56.8% of cases) or viral peptides (15.5% of cases), than their paired normal counterparts. Within our validation cohort, a linear model incorporating neoepitope binding affinity, protein similarity between neoepitopes and their closest viral peptides, and paired binding difference was able to predict actual immunogenicity with an AUROC of 0.66. Conclusions: Our proposed neoepitope prioritization criteria emphasize neoepitope novelty and refine patient neoepitope predictions for focus on biologically meaningful candidate neoantigens. We have demonstrated that neoepitopes should be considered not only with respect to their paired normal epitope, but with respect to the entire human proteome, as well as bacterial and viral peptides, with potential implications for neoepitope immunogenicity and personalized vaccines for cancer treatment. We conclude that putative neoantigens are highly variable across individuals as a function of both cancer genetics and personalized HLA repertoire, while the overall behavior of filtration criteria reflects predictable patterns.

Human leukocyte antigen susceptibility map for SARS-CoV-2

“…We kmerized each of these sequences into 8-to 12-mers to assess MHC class I-peptide binding affinity across the entire proteome. MHC class I binding affinity predictions were performed using 145 different HLA alleles for which global allele frequency data was available as described previously (72) (see Supplementary Table S1 S5) with netMHCpan v4.0 (73) using the ‘-BA’ option to include binding affinity predictions and the ‘-l’ option to specify peptides of lengths 8-12 (Supplementary Table S1). Binding affinity was not predicted for peptides containing the character ‘|’ in their sequences.…”

Burden of tumor mutations, neoepitopes, and other variants are weak predictors of cancer immunotherapy response and overall survival

“…Second, neoepitope burden was calculated for each patient weighted by amino acid mismatch as follows. The closest normal peptide in the human proteome to each neoepitope was identified using blastp (v2.6.0) [45], selecting for lowest E value or, in the case of a tie among multiple peptide sequences, the selected peptide was that with the highest weighted BLOSUM62 similarity (as described previously [46]). A neoepitope sequence was counted toward the patient's neoepitope burden once for each amino acid mismatch between the neoepitope and its closest normal peptide.…”

Tumor fitness, immune exhaustion and clinical outcomes: impact of immune checkpoint inhibitors

“…Overall our study indicates that while tumor fitness scores can be used to predict response to checkpoint inhibitors, they do not seem to predict survival in untreated patients where endogenous immune activation is a significant survival factor. It is important to note that while this tumor fitness measure and its underlying assumptions have been adapted with positive predictive results in other contexts 14,15 , we found that the exact form of the model was not predictive in these endogenous cases. Nevertheless, it is encouraging to observe that CYT measures are powerful in both treated and untreated patients, which may point to the existence of universally predictive and prognostic immune activation signatures.…”