MHCnuggets uses deep neural networks to predict neoantigen-MHC binding for both class I and II with improved accuracy for rare alleles and scalability to large datasets.
Both MHC class I & IIMHCnuggets predicts binding for both major antigen presentation pathways, unlike many tools limited to class I
What the researchers found
MHCnuggets deep neural network achieved improved prediction of MHC class I and II neoantigen binding, with better support for rare alleles and scalability to large datasets compared to existing tools.
Why it matters
Accurately predicting which neoantigens will be presented to T cells is essential for personalized cancer immunotherapy. Better predictions mean better patient selection and vaccine design.
The numbers in context
26.3M allele-peptide comparisons; <2.3 hours; 101,326 predicted IMMs; 38 hotspot genes; 24 driver genes; 4x PPV improvement
How the study worked
Deep neural network trained on MHC binding data for peptide-MHC affinity prediction. Validated against benchmark datasets. Compared performance with existing predictors for accuracy, allele coverage, and computational scalability.
Who was studied
26 cancer types from The Cancer Genome Atlas (computational analysis)
What this study cannot tell us
Computational predictions — binding prediction does not guarantee immunogenicity. Model accuracy depends on training data availability per allele. Experimental validation of predictions needed.
How to read the evidence
Moderate — validated against large benchmark datasets with demonstrated improvements, but clinical validation of predictions is ongoing.
When this study was published
Published in 2020; deep learning for neoantigen prediction continues to evolve rapidly.
The bigger picture
As cancer immunotherapy becomes standard of care, computational tools like MHCnuggets that predict neoantigen presentation are becoming essential infrastructure for precision oncology.
Questions still open
- How does MHCnuggets performance compare to newer transformer-based models?
- Can MHCnuggets predictions be integrated with T cell reactivity prediction for end-to-end neoantigen identification?
- What is the clinical impact of improved rare allele prediction on patient access to personalized vaccines?
Common questions
What is MHC binding and why does it matter for cancer treatment?
Why is rare allele support important?
Read the original research
High-Throughput Prediction of MHC Class I and II Neoantigens with MHCnuggets.
Cancer immunology research, 8(3), 396-408
Citation
Shao, Xiaoshan M; Bhattacharya, Rohit; Huang, Justin; Sivakumar, I K Ashok; Tokheim, Collin; Zheng, Lily; Hirsch, Dylan; Kaminow, Benjamin; Omdahl, Ashton; Bonsack, Maria; Riemer, Angelika B; Velculescu, Victor E; Anagnostou, Valsamo; Pagel, Kymberleigh A; Karchin, Rachel. (2020). High-Throughput Prediction of MHC Class I and II Neoantigens with MHCnuggets.. Cancer immunology research, 8(3), 396-408. https://doi.org/10.1158/2326-6066.CIR-19-0464