Human leukocyte antigen class I (HLA-I) molecules sample the proteome following the degradation of intracellular proteins via the proteasome and other proteolytic mechanisms. These complexes of HLA-I and peptide (pHLA) are then recognized by T cells and the nature of the bound peptide ligand is the key driver of adaptive immunity. The diversity of these displayed peptides, the immunopeptidome, plays an essential role in maximizing T cell responses. Recently it has been reported that a large fraction of HLA-I peptides are derived from spliced non-contiguous sequences (fusion peptides), which are not templated linearly in the genome. In order to identify such peptides, several assumptions were used that restricted potential HLA-I ligands to being derived from the same antigen and between two sequences in close proximity (proximal cis-splicing). Here we develop a sophisticated data-driven workflow to demonstrate that fusion peptides are generated through distal and proximal cis-splicing as well as trans-splicing (where distinct proteins contribute peptide segments). In addition, we find that ligation occurs predominantly between two segments of similar length and describe a new category of peptides (termed polybrid) of unknown origin. Importantly, these fusion peptides display canonical binding sequence features, although on average they had lower predicted affinity for the restricting HLA-I alleles. These results highlight the complexity and diversity of HLA-I peptide display and broaden our understanding of potential targets of T cell immunity and have important implications for vaccine design and immunotherapy.