Table 3

Predictive performance in terms of the area under the ROC curve (AUC) of the different methods evaluated on six data sets.

AntiJen

IEDB


DRB1*0101

DRB1*0401

DRB1*1501

DRB1*0101

DRB1*0401

DRB1*1501


ISC-PLC

0.709

0.757

0.609

SMM-align

0.718

0.806

0.691

0.702

0.741

0.715

TEPITOPE

0.667

0.744

0.665

0.647

0.754

0.726

Chang

0.770

0.757

0.677

SMM-regr

0.807

0.819

0.741

0.744

0.750

0.718

SMM-regr-alter

0.616

0.785

0.669

0.645

0.721

0.712

SMM-PFR

0.742

0.814

0.726

0.716

0.756

0.733


The methods are; ISC-PLS [15], SMM-align, TEPITOPE, Chang [11], SMM-regr (SMM with peptide length regression correction from training data set), SMM-regr-alter (SMM with peptide length regression correction from alternative AntiJen/IEDB dataset), and SMM-PFR (The SMM-PRF method refers to the extended SMM align method including penalties for long peptides and short amino terminal peptide flanking residues). The data sets consist of peptides binding data from two sources (IEDB and AntiJen) covering three HLA-DR alleles (1*0101, 1*0401, and 1*1501). Performance value for the ISC-PLS and Chang methods, are taken from Chang et al. [11]. These values are only available for the AntiJen data set.

Nielsen et al. BMC Bioinformatics 2007 8:238   doi:10.1186/1471-2105-8-238

Open Data