For many common cancers, survival and efficacy of state-of-the-art therapies are strongly predicted by the number of non-synonymous mutations. It is widely hypothesized that this association between mutations and survival is a result of mutations leading to neo-epitopes which in turn induce a robust anti-tumor immune response. It follows that accurately quantifying the number of neo-epitopes within a patient should better predict survival than that of non-synonymous mutations, however that has generally not been the case. Indeed, mutations and predicted neo-epitopes explain roughly the same survival variance, likely due to the difficulty of accurately predicting and prioritizing immunogenic neo-epitopes from mutations. We hypothesized that by integrating several data types and using state-of-the-art techniques for major histocompatibility complex (MHC) class I and II allele calling and epitope binding prediction, we could predict survival of cancer patients better than with mutation calls alone. We have developed a pipeline to characterize the potential immunogenicity of neo-epitopes. The pipeline uses copy-number variants and reads from whole-exome sequencing to assess the clonality of a given neo-epitope. MHC class I and class II allelotyping is performed, and expression of MHC alleles and mutations is determined using RNA-Seq. Similarity of epitopes to self-peptides is assessed by measuring similarity to predicted self-ligands, as well as by measuring the MHC affinity of wild-type versions of mutated epitopes. With this pipeline we were able to rapidly process the cancer genome atlas (TCGA) breast cancer (BRCA), non-small cell lung cancer (NSCLC), melanoma (SKCM), and pancreatic cancer (PAAD) datasets to infer neo-epitopes burden with high resolution. A linear model of survival as a function of mutations and their putative immunogenicity was constructed using these datasets via elastic net regularization, while modeling variance was controlled with k-fold cross validation. Our results confirm that the putatively immunogenic mutational burden is a highly significant term is such models. Finally, we quantify epitope response in these models by incorporating TCR sequence entropy as a proxy for clonal tumor infiltrating lymphocyte expansion.
Citation Format: Nicholas K. Akers, Eric Schadt, Bojan Losic. Predicting cancer survival with neo-epitope burden [abstract]. In: Proceedings of the Second CRI-CIMT-EATI-AACR International Cancer Immunotherapy Conference: Translating Science into Survival; 2016 Sept 25-28; New York, NY. Philadelphia (PA): AACR; Cancer Immunol Res 2016;4(11 Suppl):Abstract nr A024.
- ©2016 American Association for Cancer Research.