We provide predictions for USMPep on the following datasets:
- HPV by Bonsack2019: Bonsack M, Hoppe S, Winter J, Tichy D, Zeller C, Küpper MD, et al., Performance evaluation of MHC class-I binding prediction tools based on an experimentally validated MHC–peptide binding data set. Cancer Immunol Res , vol. 7, no. 5, pp. 719–736, Mar. 2019, https://doi.org/10.1158/2326-6066.CIR-18-0584
- IEDB16_I and IEDB16_II by Zhao W, Sher X, Systematically benchmarking peptide-MHC binding predictors: From synthetic to naturally processed epitopes. PLoS Comput Biol, vol. 14, no. 11, p. e1006457, Nov. 2018, https://doi.org/10.1371/journal.pcbi.1006457
- Kim14 by Kim Y, Sidney J, Buus S, Sette A, Nielsen M, Peters B, Dataset size and composition impact the reliability of performance benchmarks for peptide-MHC binding predictions. BMC Bioinformatics, vol. 15, no. 1, p. 241, 2014. https://doi.org/10.1186/1471-2105-15-241
In addition to IC50 predictions (*_predictions.csv files), we provide the raw predictions which are the binding affinities transformed to values between 0.0 (non-binder) and 1.0 (binder) in the *_predictions_log.csv files.