Background: The identification of broadly shared, highly immunogenic tumor antigens remains a central challenge in cancer vaccine development. Non-canonical tumor peptides (nonC-TPs) arising from the cancer “dark genome” have recently emerged as a highly immunogenic class of tumor antigens with strong potential for off-the-shelf cancer vaccine development. These peptides originate from aberrant transcriptional and translational events involving transposable elements, endogenous retroviruses, untranslated regions, and alternative open reading frames, represent a promising antigen class, as they are absent from normal tissues, largely excluded from thymic selection, and shared across tumor histologies and patients. Molecular mimicry between tumor antigens and viral epitopes has been previously demonstrated for canonical tumor-associated antigens (TAAs), providing a mechanistic basis for cross-reactive anti-tumor immunity mediated by virus-specific memory T cells. This has never been investigated in the nonC-TP landscape of the dark genome. Methods: Starting from the proteogenomic dataset generated by Lozano-Rabella, comprising 65,504 canonical and non-canonical peptides identified in several primary cancer cell lines, we prioritized 244 tumor-exclusive non-canonical nonamers predicted as strong HLA binders using NetMHCpan4.1. Then we focused our analysis on restricted NonC peptides binding four HLA-A allotypes. Homology searches against viral proteomes were performed using BLAST, followed by HLA binding and stability prediction with NetMHCstabpan1.0. Structural modelling and molecular docking analyses were conducted to evaluate conservation of peptide-HLA molecular conformations between nonC peptides and viral homologs. Results: A significant subset of dark genome-derived nonC TPs displayed significant sequence homology with viral epitopes across multiple HLA-A alleles, particularly HLA-A*02:01, HLA-A*03:01, and HLA-A*11:01. Homologous viral sequences were identified from clinically relevant pathogens. The homologous peptides frequently conserved amino acid residues predicted to mediate TCR interaction, supporting the possibility of cross-reactive recognition by pre-existing anti-viral memory CD8+ T cells. Structural modelling further demonstrated highly similar peptide-HLA docking conformations between nonC peptides and their viral counterparts, supporting the potential role of molecular mimicry within the dark immunopeptidome. Conclusions: This study provides the first systematic evidence that dark genome-derived nonC-TPs exhibit molecular mimicry with viral antigens, extending a well-established immunological paradigm to a previously unexplored antigen class. The identified viral-homologous nonC-TPs combine tumor specificity, broad sharing across patients and histologies, and potential recognition by pre-existing virus-specific memory T cells, positioning them as uniquely attractive candidates for next-generation, off-the-shelf cancer vaccines. These findings open a new dimension of tumor immunosurveillance and provide a rational framework for the development of immunotherapeutic strategies that exploit the convergence of dark genome biology, molecular mimicry, and viral mimicry.



