TIPs: a deep learning-guided proteogenomic framework to expand the landscape of transposable element-derived antigens with immunopeptidomics.

Transposable elements (TEs) represent an abundant and important source of HLA-presented antigens, but their immunopeptidomic characterization remains challenging due to the inflated search space. We present TIPs (TE-derived Immunopeptidomic Search), a deep learning-guided proteogenomic framework that integrates de novo sequencing, database refinement, multiple search engines and stringent FDR controls. Across various cell lines and cancer types, TIPs identified 20-fold more TE-derived peptides on average than conventional approaches. It further revealed many recurrent, tumor-specific antigens from TEs, including candidates induced by epigenetic therapy. These findings highlight the potential of TIPs to expand the antigenic landscape beyond canonical sources.