MiHAs with polymorphic AAs at nonanchor positions are predicted to bind similarly to HLA as their allelic variants. MiHAs binding to different HLA-I alleles with polymorphic AAs at anchor or nonanchor positions were compared for predicted HLA-binding. (A) The 81 new MiHAs (orange) that were identified by optimized GWAS were mainly found for HLAs for which the method was optimized (HLA-A∗01:01, A∗02:01, A∗03:01, B∗07:02, B∗08:01, C∗07:01, C∗07:02), but also HLAs that were expressed by sufficient numbers of EBV-LCLs in the GWAS panel (HLA-A∗24:02, B∗35:01, B∗40:01, B∗44:02, B∗44:03, C∗03:03) or HLAs that were retrovirally introduced in EBV-LCLs of the GWAS panel (A∗68:01). MiHAs that were previously identified (blue) are also shown. (B) All 142 MiHAs with a single polymorphic AA were divided into 2 groups based on polymorphic residues being at an anchor position (blue) or not (gray). With 20 (80.0%) of 25 MiHAs with a polymorphic anchor residue, the majority of MiHAs showed at least 5 times stronger predicted HLA-binding (outside light gray background) than their allelic counterpart compared to only 13 (11.5%) of 113 MiHAs with a nonanchor polymorphic AA. Indicated are predicted HLA-binding scores by NetMHCpan4.1 for MiHAs (x-axis) and allelic variants (y-axis). High % rank values indicate low predicted HLA-binding. Red and blue dotted lines indicate the cut-offs for strong (% rank <0.5) and weak (% rank <2.0) predicted HLA-binding, respectively.