The abstract is read at the publisher; the summary is JClub's.
DeepOmicsNeo improves non-9mer MHC class I binding and overall neoantigen immunogenicity prediction, achieving a drastically higher Top20 precision on TESLA, HiTIDE, and TiTANIAN benchmarks.
- Why it matters: Personalized cancer vaccine design depends on accurate neoantigen identification, but existing predictors generate many false positives, limiting clinical utility.
- What they did: DeepOmicsNeo combines three modules: DeepOmicsNeo-Binding uses pseudo-binder/non-binder augmentation by editing non-anchor residues; DeepOmicsNeo-Selfness trains a GAN on ~60 million human self-peptides and links its selfness score to IFN-gamma secretion; DeepOmicsNeo-H2 models central-region homology to class II epitopes.
- The result: The integrated DeepOmicsNeo-Immunogenicity outperforms state-of-the-art predictors on TESLA, HiTIDE, and TiTANIAN, with Top20 precision most strongly improved, supporting neoantigen prioritization for personalized cancer immunotherapy.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.