Robust prediction of HLA class II epitopes by deep motif deconvolution of immunopeptidomes

Article metrics


Predictions of epitopes presented by class II human leukocyte antigen molecules (HLA-II) have limited accuracy, restricting vaccine and therapy design. Here we combined unbiased mass spectrometry with a motif deconvolution algorithm to profile and analyze a total of 99,265 unique peptides eluted from HLA-II molecules. We then trained an epitope prediction algorithm with these data and improved prediction of pathogen and tumor-associated class II neoepitopes.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.


All prices are NET prices.

Fig. 1: Motif deconvolution in HLA-II peptidomics data.
Fig. 2: MixMHC2pred improves class II epitope prediction.

Data availability

The raw datasets generated during the current study are available in the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD012308 and the corresponding peptide output files are provided as Supplementary Data 1 and 2. Additional datasets generated during the current study are available as Supplementary Tables 1 and 4. In addition, public datasets were analyzed in this study, obtained from the IEDB database19 and from the studies listed in Supplementary Table 2, as well as from multiple neoantigen studies listed in Supplementary Data 3.

Code availability

MoDec and MixMHC2pred are freely available as C++ executables ( and Supplementary Code 1 and 2) for academic non-commercial research purposes. MixMHC2pred is also freely available for academic non-commercial research purposes as a web application (


  1. 1.

    Neefjes, J., Jongsma, M. L. M., Paul, P. & Bakke, O. Nat. Rev. Immunol. 11, 823–836 (2011).

  2. 2.

    Khodadoust, M. S. et al. Nature 543, 723–727 (2017).

  3. 3.

    Linnemann, C. et al. Nat. Med. 21, 81–85 (2015).

  4. 4.

    Kreiter, S. et al. Nature 520, 692–696 (2015).

  5. 5.

    Andreatta, M. et al. Immunogenetics 67, 641–650 (2015).

  6. 6.

    Andreatta, M. et al. Bioinformatics 34, 1522–1528 (2018).

  7. 7.

    Chong, C. et al. Mol. Cell. Proteomics 17, 533–548 (2018).

  8. 8.

    Ritz, D. et al. Proteomics 18, 1700246 (2018).

  9. 9.

    Bassani-Sternberg, M. et al. Nat. Commun. 7, 13404 (2016).

  10. 10.

    Bassani-Sternberg, M. & Gfeller, D. J. Immunol. 197, 2492–2499 (2016).

  11. 11.

    Bassani-Sternberg, M. et al. PLoS Comput. Biol. 13, e1005725 (2017).

  12. 12.

    Abelin, J. G. et al. Immunity 46, 315–326 (2017).

  13. 13.

    Jurtz, V. et al. J. Immunol. 199, 3360–3368 (2017).

  14. 14.

    Bulik-Sullivan, B. et al. Nat. Biotechnol. 37, 55–63 (2019).

  15. 15.

    Barra, C. et al. Genome Med. 10, 84 (2018).

  16. 16.

    Garde, C. et al. Immunogenetics 71, 445–454 (2019).

  17. 17.

    Nielsen, M. & Andreatta, M. Nucleic Acids Res. 45, W344–W349 (2017).

  18. 18.

    Andreatta, M., Alvarez, B. & Nielsen, M. Nucleic Acids Res. 45, W458–W463 (2017).

  19. 19.

    Vita, R. et al. Nucleic Acids Res. 47, D339–D343 (2019).

  20. 20.

    Jensen, K. K. et al. Immunology 154, 394–406 (2018).

  21. 21.

    Bailey, T. L. & Elkan, C. Fitting a mixture model by expectation maximization to discover motifs in biopolymers. in Proc. Second International Conference on Intelligent Systems for Molecular Biology (eds Altman, R., Brutlag, D., Karp, P., Lathrop, R., & Searls, D.) 28–36 (AAAI, 1994).

  22. 22.

    Falk, K., Rötzschke, O., Stevanovíc, S., Jung, G. & Rammensee, H.-G. Immunogenetics 39, 230–242 (1994).

  23. 23.

    Ciudad, M. T. et al. J. Leukoc. Biol. 101, 15–27 (2017).

  24. 24.

    Gfeller, D. et al. J. Immunol. 201, 3705–3716 (2018).

  25. 25.

    Yin, L., Calvo-Calle, J. M., Dominguez-Amorocho, O. & Stern, L. J. J. Immunol. 189, 3983–3994 (2012).

  26. 26.

    Dudley, M. E. et al. Clin. Cancer Res. 16, 6122–6131 (2010).

  27. 27.

    Donia, M., Larsen, S. M., Met, Ö. & Svane, I. M. Cytotherapy 16, 1117–1120 (2014).

  28. 28.

    Vizcaíno, J. A. et al. Nucleic Acids Res. 44, D447–D456 (2016).

  29. 29.

    Cox, J. & Mann, M. Nat. Biotechnol. 26, 1367–1372 (2008).

  30. 30.

    Gfeller, D. et al. Mol. Syst. Biol. 7, 484 (2011).

  31. 31.

    Wagih, O. Bioinformatics 33, 3645–3647 (2017).

  32. 32.

    Nielsen, M. et al. Bioinformatics 20, 1388–1397 (2004).

  33. 33.

    Clement, C. C. et al. J. Biol. Chem. 291, 5576–5595 (2016).

  34. 34.

    Collado, J. A. et al. Eur. J. Immunol. 43, 2273–2282 (2013).

  35. 35.

    Ooi, J. D. et al. Nature 545, 243–247 (2017).

  36. 36.

    Wang, Q. et al. J. Proteome Res. 16, 122–136 (2017).

  37. 37.

    Bergseng, E. et al. Immunogenetics 67, 73–84 (2015).

  38. 38.

    Bobisse, S. et al. Nat. Commun. 9, 1092 (2018).

  39. 39.

    Chattopadhyay, P. K., Yu, J. & Roederer, M. Nat. Protoc. 1, 1–6 (2006).

  40. 40.

    Ott, P. A. et al. Nature 547, 217–221 (2017).

  41. 41.

    Tran, E. et al. Science 350, 1387–1390 (2015).

  42. 42.

    Veatch, J. R. et al. J. Clin. Invest. 128, 1563–1568 (2018).

  43. 43.

    Veatch, J. R et al. Cancer Immunol. Res. 7, 910–922 (2019).

  44. 44.

    Yossef, R et al. JCI Insight 3, e122467 (2018).

  45. 45.

    Zacharakis, N. et al. Nat. Med. 24, 724–730 (2018).

  46. 46.

    Sahin, U. et al. Nature 547, 222–226 (2017).

Download references


We thank the Center of Experimental Therapeutics team for providing us with the patient-derived tissue samples and T cells. We thank P. Romero from the University of Lausanne for sharing the B cell lines with us. We thank R. T. Daniel and M. Hegi from the University Hospital of Lausanne for providing us with the collection of meningioma tissues. We thank M. Solleder for help with the visualization of motifs with ggseqlogo, F. Marino for technical support with sample preparation, H.-S. Pak for MS measurements and R. Genolet for HLA typing. This work was supported by the Swiss Cancer League (grant KFS-4104-02-2017 to D.G. and J.R.), the Ludwig Institute for Cancer Research, the ISREC Foundation thanks to a donation from the Biltema Foundation (to J.M., C.C. and M.B.-S.) and by the MEDIC foundation (to G.A.R. and C.J.).

Author information

J.R. developed the computational methods; J.R. and D.G. analyzed the data; J.M., C.C. and M.B.-S. generated the MS peptidomics data; G.A.R., M.A., S.B., P.G., A.H. and C.J. performed the binding and T cell assays; G.C., A.H., C.J. and M.B.-S. provided reagents; J.R., M.B.-S. and D.G. designed the study; and J.R., M.B.S. and D.G. wrote the paper.

Correspondence to Michal Bassani-Sternberg or David Gfeller.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Supplementary Figs. 1–16, Supplementary Tables 1–3 and Supplementary Note.

Reporting Summary

Supplementary Table 4

Viral, bacterial and tumor-associated epitopes. aList and full sequence of the viral, bacterial and tumor-associated antigens included in the experiment. bBest scoring candidate epitopes predicted by MixMHC2pred or NetMHCIIpan that were tested for immunogenicity in two patients with melanoma (LAU1352 and LAU1357) and a healthy donor.

Supplementary Data 1

List of peptides identified in the pan-HLA-II peptidomics data. For the JY cell line, an experiment with an anti-HLA-DR antibody was also performed in the same runs and is included in this table (labeled JY_DR).

Supplementary Data 2

List of peptides identified in the HLA-DR and HLA-DR-depleted peptidomics data.

Supplementary Data 3

Benchmark data containing neoepitopes from various studies. aList of peptides tested experimentally indicating which ones were CD4+ T cell immunogenic. bHLA typing from each patient and number of positive and negative epitopes tested experimentally.

Supplementary Code 1

MoDec. Executable from MoDec (v.1.1) performing the motif deconvolution.

Supplementary Code 2

MixMHC2pred. Executable from MixMHC2pred (v.1.1), predicting HLA-II ligands and epitopes.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Racle, J., Michaux, J., Rockinger, G.A. et al. Robust prediction of HLA class II epitopes by deep motif deconvolution of immunopeptidomes. Nat Biotechnol 37, 1283–1286 (2019) doi:10.1038/s41587-019-0289-6

Download citation