Prediction of Splicing-Associated Neoepitopes

An interesting new research direction, which grew out of our cancer transcriptomics projects, is the prediction of cancer-specific splicing-derived neoepitopes. Neoepitopes are at the core of modern Immunotherapies. Like how somatic mutations can generate specific peptide signatures that occur uniquely in the tumour but not in the normal cells, aberrant splicing can generate novel exon-exon junctions that translate into tumour-specific amino acid sequences.

We are developing the ImmunoPepper software, which uses a comprehensive splicing graph built on foreground and background cohorts to predict putative tumour-specific neoepitopes.  ImmunoPepper allows for the generation of mutation and splicing-derived peptides in both cohorts. It also enables several filtering steps which help increase confidence in the candidates while minimising the risk of cancer vaccine adverse effects on normal tissues. Together with additional data sources, such as mass spectrometry datasets, we can further filter and validate the predicted candidates at the protein level.

Finally, we aim to make splicing-derived neoepitope discovery a reproducible and robust analysis. We collaborate closely with others developing junction-centric basic approaches to neoepitope prediction. With the comparison of our graph-centric and the junction-centric approach, we plan to provide the immuno-oncology community with a clear roadmap to strategies for deriving neoepitopes.

The ImmunoPepper software is available as open source software. [Download]

Involved group members: Laurie Prelot, Andre Kahles, Matthias Hüser (alumnus), Gunnar Rätsch