7/ ii) Binding affinity predictions don’t always correlate with actual presentation at the cell surface. More immunopeptidomics is essential—not just to rely on binding predictions but to better assess HLA presentation.

6/ Two additional (and not-so-original) thoughts:
i) Mutations generate the best neoantigens, but each patient may require more antigens for an effective vaccine. --> additional sources (cryptic proteome or alternative splicing) may help?

5/ Both findings suggest that, beyond lower binding affinity, non-responder peptides are also less likely to be presented.

4/ Also, Müller et al.(PMID:29104575) proposed that if the WT peptide is presented, the mutated version is more likely to be presented. We searched for WT counterpart peptides in an in-house dataset of 1000s of immunopept. samples and found a higher proportion of presented peptides in responders.

3/ However, HLA-I binding affinity doesn’t guarantee real peptide presentation. I explored PDAC neoantigen presentation using two approaches.
First, Cysteines are generally depleted in the immunopeptidome. Notably, non-responder peptides had a 3-fold increase in cysteine-containing peptides.

2/ Responder peptides had a higher HLA-I affinity than those in non-responders (as expected). ~50% of the non-immunogenic peptides weren’t binders (by NetMHC). This suggests that one distinction between responders and non-responders lies in the quality of the neoantigens included in the vaccine.