To tap on our adaptive immunity for cancer therapy, knowledge of tumor-specific peptide antigens is crucial. As such, the race is on from the proteogenomics, mass spectrometry, immunopeptidomics and cancer immunology fronts, to find key antigenic peptide sequences, that are ideally prevalent, highly conserved across different tumors and promiscuous in loading to different patient HLA types. In our view, the bottleneck to achieve this is no longer sequence-based variant calling, given the exciting progress of computational approaches and database assembly in the recent years. Instead, we believe by tracing the steps of antigen processing, we can assemble the working rules of what will become a good antigen, and what will not. To this end, we have added protein-centric filters to prioritise antigen targets on top of mutation-based antigen shortlisting, and reveal additional mechanisms of escape from antigen presentation via exosome shedding. Collectively, these can strongly impact the utility of candidate peptides that emerge out of sequence-based antigen discovery.