Skip to main content

Table 3 Cross-validation performance for the combined NNAlign models, trained on both binding affinity (BA) and eluted ligand (EL) data

From: Footprints of antigen processing boost MHC class II natural ligand predictions

Training set

BA prediction

EL prediction

BA

EL

AUC

AUC 0.1

AUC

AUC 0.1

PPV

DR1 BA

DR1 Ph

0.845

0.385

0.966

0.823

0.781

DR1 Pm

0.843

0.376

0.987

0.893

0.826

DR1 Sm

0.843

0.381

0.98

0.867

0.814

DR15 BA

DR15 Ph

0.844

0.288

0.987

0.908

0.855

DR15 Pm

0.846

0.294

0.99

0.917

0.86

DR51 BA

DR51 Ph

0.848

0.389

0.956

0.749

0.74

  1. Training set refers to the data set used to train the given model (BA indicated binding affinity and EL eluted ligand data). For references on the training sets names and compositions, refer to Table 1. Cross-validated performance values are reported as AUC, AUC 0.1, and PPV. For more details on these measures, refer to the “Methods” section. Note that minor variations in the BA performance values for the same molecule are due to the differences in the data partitioning in the fivefold cross-validation setup in each case