DR-341 Characterization of immunogenicity of mutations

Some tumor show spectacular effect of immunotherapy, while other tumors show only limited response. Finding biomarkers to predict this response, in order to more carefully select patients for these expensive treatments, has been a major effort, resulting in many suboptimal biomarkers (different clones of anti-PD-L1, tumor mutational burden, etc.). It has been shown that individual mutations can be more or less immunogenic. However, individual mutations can also be classified as ‘more immunogenic’ and ‘less immunogenic’ (Marty R. et al., Cell, 2017). We propose a more in-depth classification of mutations, using a more unbiased approach, using machine learning models to process amino acids in peptides. We hope using the HMF databases of mutations and germline data for MHC classes, to make a better model for predicting immunogenicity of tumors.

Michaël Noë, Stichting Het Nederlands Kanker Instituut – Antoni van Leeuwenhoek Ziekenhuis, the Netherlands