Abstract 567: Comparative mining of normal and tumor tissue RNAseq gene expression datasets to define expression-based neo-antigens

Abstract

Background: Mutational neo-antigens (MuNeoAgs) are attractive targets for cancer vaccines because they are unique to tumors. However, limited epitope potential of point mutations, inherent difficulties predicting immunogenicity, and lack of shared mutations across patient tumors has hindered vaccine development. Alternatively, we elected to identify non-mutated, expression-based tumor-specific neo-antigens (EbNeoAgs) defined by their overexpression in tumors and negligible expression in healthy tissue. EbNeoAgs have a low probability of cross-reactivity, while allowing epitope creation across the entire protein. Additionally, they should be present in a higher proportion of patients and could be used to make modular off-the-shelf vaccines. Methods: RNAseq data from the NIH Genotype-Tissue Expression (GTEx) and The Cancer Genome Atlas (TCGA) projects, previously harmonized with a common calling/ mapping algorithm and available from the University of California Santa Cruz TOIL RNAseq recompute project, were used to calculate upper quartile FPKM gene expression values for all samples based on protein coding RNAs. RNAseq data from 27 normal tissue types (excluding testis) from GTEx and TCGA normal tissue was analyzed to identify genes with negligible expression across all healthy tissue. Genes with negligible expression were then analyzed to determine if they were elevated (i.e. expressed at intermediate or higher levels) in samples across 32 different cancer types (excluding testicular tumors). Thresholds for intermediate expression were calculated individually for each tumor type based upon the median FPKM expression values for all protein coding genes with 95% confidence intervals. Results: 1296 genes were expressed below threshold in all healthy tissues, and 107 of these genes were elevated in 5% or more of samples in at least one cancer. 40/107 EbNeoAgs (37%) are known Cancer Testis Antigens (CTAs), with the remaining 67 representing potentially novel EbNeoAgs. If the threshold for elevated expression is lowered to just 1% of samples for at least one cancer, 58 of the known 276 CTAs are detected. The remaining CTAs are overexpressed in at least one healthy tissue or were not elevated in cancer. In melanoma (cancer with the most EbNeoAgs), 53 (52.0%), 69 (67.6%), and 88 (86.3%) samples expressed at least 5, 3, or 1 EbNeoAg respectively. Comparatively, in lung adenocarcinoma, 81 (15.8%), 97 (18.9%), and 228 (44.4%) samples expressed at least 5, 3, or 1 EbNeoAg respectively. Conclusion: We leveraged publicly available tumor and normal RNAseq datasets to comprehensively identify EbNeoAgs in an unbiased fashion across 32 tumor types. Many of the EbNeoAgs were known CTAs, but a significant number of EbNeoAgs were novel. Approximately half the cancers exhibited EbNeoAg expression in a large proportion of patients making them broad targets for cancer vaccine development. Citation Format: Jan O. Kemnade, Maria Cardenas, David Wheeler, Andrew G. Sikora, Mitchell J. Frederick. Comparative mining of normal and tumor tissue RNAseq gene expression datasets to define expression-based neo-antigens [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 567.