NetMHCIIpan-2.0 - Improved pan-specific HLA-DR predictions using a novel concurrent alignment and weight optimization training procedure

Table 7 Predictive performance in terms of the AUC on the Lin benchmark data set

Allele	*NetMHCIIpan-2.0*	*NetMHCIIpan-1.0*	*TEPITOPE*	Multipred_SVM	SVMHC
DRB1*0101	0.883	0.847	0.919	0.860	0.860
DRB1*0301	0.716	0.668	0.718	0.800	0.690
DRB1*0401	0.846	0.815	0.745	0.650	0.750
DRB1*0701	0.878	0.852	0.715	0.700	0.740
DRB1*1101	0.884	0.821	0.824	0.780	0.830
DRB1*1301	0.729	0.715	0.718	0.630	0.720
DRB1*1501	0.838	0.791	0.737	0.620	0.660
Ave	0.825	0.787	0.768	0.720	0.750

The AUC was calculated using the following binding affinity threshold values for each of the 7 alleles: DRB1*0101, 0401, 0701, and 1501 threshold = 100 nM, DRB1*0301, 1101, and 1301, threshold = 1000 nM. The performance values for Multipred_SVM and SVMHC were taken from Nielsen et al. [5]. TEPITOPE is the method described by Sturniolo et al. [1]. NetMHCIIpan-2.0 is the pan-specific method described here, and NetMHCIIpan-1.0 is the pan-specific method by Nielsen et al. [29]. For each allele, the best performing method is highlighted in bold.