1204 Repetitive element-derived neoantigens are a potential source of highly tumor-specific cancer vaccine targets

<h3>Background</h3> Repetitive elements are suppressed in normal tissues but commonly expressed in cancer, where methylation changes can lead to their reactivation. Combined with dysregulation of splicing machinery, this leads to an increase in the number of RNA repeat-gene ‘fusions’: transcripts derived from junctions between protein-coding genes and repetitive elements. These fusions have the potential to be a rich source of highly tumor-specific cancer vaccine targets. <h3>Methods</h3> ROME has developed a state-of-the-art ‘repeatomics’ machine learning platform designed to quantify the expression of highly repetitive elements with substantially improved sensitivity and specificity. This platform includes ‘ROMEFuse™’, which integrates short-read assembly and an alignment-based approach to detect RNA fusions between genes and repeat elements. We used ROMEFuse to profile the landscape of repeat-gene fusions in large-scale, publicly available tumor and normal RNA-seq data (TCGA, GTEx), and additionally in a set of esophageal tumor, normal-adjacent to tumor (NAT), and normal esophagus samples (N = 5 each; N = 15 total). We also re-analyzed public immunopeptidomics data to confirm that fusion peptides can be presented by diverse MHC class I molecules. <h3>Results</h3> Using ROMEFuse, we identified multiple classes of repeat fusions prevalent in both TCGA and in-house acquired tumor samples, including two classes likely to result in the production of novel protein products: fusions originating from genes and other fusions originating from an antisense promoter within LINE-1 elements. A subset of these is highly tumor-specific across multiple cancer indications and found with high prevalence in the diseased population. These fusions involve genes previously reported in fusion events as well as newly implicated ones. We also find significant heterogeneity in fusion prevalence by indication, with some indications harboring fusions detected across many genes while others harbor substantially fewer. Of the tumor specific fusions identified, over 20 peptides originating from these fusions were also detected in public immunopeptidomics data, some of which overlap the junctions between genes and repeats. <h3>Conclusions</h3> Repetitive elements have been shown to be highly expressed in cancer. However, repeat-gene RNA fusions are less well studied due to the lack of efficient and accurate tools. Here, we show our novel algorithm, ROMEFuse, can identify a subset of fusion-derived peptides that are tumor-specific and detected in publicly available immunopeptidomics data. Given the potentially large space of repeat-gene fusions, their tumor specificity, and patient prevalence, we believe that these elements could be a rich source of cancer vaccine targets.