Abstract
Using personalized peptide vaccines (PPVs) to target tumor-specific non-self antigens (neoantigens) is a promising approach to cancer treatment. However, the development of PPVs is hindered by the challenge of identifying tumor-specific neoantigens, in part because current in silico methods for identifying such neoantigens have limited effectiveness. Here we report the results of molecular dynamics simulations of 12 oligopeptides bound with a human leukocyte antigen (HLA), revealing a previously unrecognized association between the inability of an oligopeptide to elicit a T-cell response and the contraction of the peptide-binding groove upon binding of the oligopeptide to the HLA. Our conformational analysis showed that this association was due to incompatibility at the interface between the contracted groove and its αβ–T-cell antigen receptor (TCR). This structural demonstration that having the capability to bind HLA does not guarantee immunogenicity prompted us to develop an atom-based method to predict immunogenicity through using the structure and energy of a peptide•HLA complex to assess the propensity of the complex for forming a ternary complex with its TCR. In predicting the immunogenicities of the 12 oligopeptides, achieved a 100% success rate compared with success rates of 25–50% for 11 publicly available residue-based methods including NetMHC-4.0. While further validation and refinements of are required, our results suggest a need to develop in silico methods that assess peptide characteristics beyond their capability to form stable binary complexes with HLAs to help remove hurdles in using the patient tumor DNA information to develop PPVs for personalized cancer immunotherapy.
Publisher
Cold Spring Harbor Laboratory