MPN driver mutations (CALR and MPL) and SF3B1-mediated splicing aberrations are a rich source of predicted neoantigens. (A) Number of predicted neoantigens (nonsynonymous SNVs, indels, and fusions) for each patient, based on individual HLA haplotypes. For prediction, the 4 most common alleles for HLA-A, HLA-B, and HLA-C within the MPN cohort were considered (supplemental Table 19). Thresholds for weak binders (%Rank < 2 and >0.5) and strong binders (%Rank < 0.5) were selected based on the NetMHCpan authors’ recommendation. Percentile rank score (%Rank) values for CALR and MPL (B) and SF3B1 peptides predicted with NetMHCpan (C) are plotted. HLA alleles, marked in red, were used for validation (Figure 6). HLA allele frequencies for the MPN cohort are indicated in panel C. (D) Number of predicted neoantigens (all mutation classes combined) for each patient with no CALR or SF3B1 mutation, 1 of the 2 mutations, or both mutations, separated into weak and strong binders. CALR-mutated patients had a mean of 8.0 weak binders and 2.3 strong binders, whereas SF3B1-mutated patients, without a cooccurring CALR mutation, had an average of 38.2 weak binders and 16.2 strong binders. For JAK2-V617F–mutated patients, we found an average of 0.9 and 0.1 predicted neoantigens for weak and strong binders, respectively (supplemental Table 23). (E) CALR mutant tail “consensus sequence” as reported by Klampfl et al.¹  The dots indicate the starting positions of peptides colored by peptide length.