Fig. 2From: Predicting peptide presentation by major histocompatibility complex class I: an improved machine learning approach to the immunopeptidomeRF-based method outperforms existing predictors on unbalanced data. a AUPRC and b Prec1% on withheld test dataare greater for RF compared to NetMHC, NetMHCpan, and NetMHCstabpan, with no significant difference between MixMHCpred and RF (p > 0.01). P values are by Mann-Whitney U Test compared to ForestMHC. Data for this figure are provided in Additional file 4Back to article page