Machine learning improves prediction of immunogenic neoantigens

Researchers from the UNIL-CHUV Department of Oncology and the Lausanne branch of the Ludwig Institute publish a new study in "Immunity", demonstrating that machine learning improves the prioritization of specific mutations in certain cancer candidates for immunotherapy treatment.

Some forms of cancer contain numerous mutations, but only a restricted sub-population, usually a few dozen, can be incorporated into the immunotherapy strategy. So the crucial step in the immunotherapy process is to carefully select the mutations with the greatest potential for successful immunotherapy.

The study * led by Markus Müller and directed by Michal Bassani-Sternberg, analyzed sequencing data from around 100 cancer patients. Going beyond conventional neoantigen prioritization features, the study identified key determinants such as neopeptide location in HLA presentation hotspots, binding promiscuity and oncogenicity of the mutated gene, providing crucial information for predicting immunogenicity.

By integrating machine learning methods, the team developed classifiers that accurately predicted neoantigen immunogenicity in diverse datasets, improving neoantigen classification by 30%. In addition, the research resulted in valuable standardized datasets to advance and evaluate associated algorithms in neoantigen-based immunotherapies.

This groundbreaking research opens up new perspectives for immunotherapies by providing an in-depth understanding of neoantigen selection and its impact on immunogenicity, ultimately aiming to revolutionize cancer treatment strategies.

* Machine learning methods and harmonized datasets improve immunogenic neoantigen prediction