The primary manuscript describing the mhc_epitope scoreterm is currently (June 2019) in preparation.  This information will be updated when it is published.

The Predictor used in de-immunization in this scientific test is Propred, and is decribed here:
Singh H, Raghava GPS. ProPred: prediction of HLA-DR binding sites. Bioinformatics 2001;17:1236–7. https://doi.org/10.1093/bioinformatics/17.12.1236.

The benchmark set, and a previous implementation of a packer-compatible de-immunization scoreterm, is described in the following article:
Choi Y, Griswold KE, Bailey-Kellogg C. Structure-based redesign of proteins for minimal T-cell epitope content. J Comput Chem 2013;34:879–91. https://doi.org/10.1002/jcc.23213.

The "nmer" SVM-based method of de-immunizing using Rosetta is NOT packer compatible in its original form, but can now be used with mhc_epitope in a packer-compatible manner.  It is NOT used in this benchmark.  It is described here:
King C, Garza EN, Mazor R, Linehan JL, Pastan I, Pepper M, et al. Removing T-cell epitopes with computational protein design. Proc Natl Acad Sci U S A 2014;111:8577–82. https://doi.org/10.1073/pnas.1321126111.

Additional Predictors, not written for Rosetta but accessible through mhc_epitope, are described below.  Note that these are NOT used in this benchmark.
Jensen KK, Andreatta M, Marcatili P, Buus S, Greenbaum JA, Yan Z, et al. Improved methods for predicting peptide binding affinity to MHC class II molecules. Immunology 2018;154:394–406. https://doi.org/10.1111/imm.12889.

Vita R, Mahajan S, Overton JA, Dhanda SK, Martini S, Cantrell JR, et al. The Immune Epitope Database (IEDB): 2018 update. Nucleic Acids Res 2019;47:D339–43. https://doi.org/10.1093/nar/gky1006.
