Predicting Class II MHC-Peptide binding: a kernel based approach using similarity scores

BMC Bioinformatics

Table 3 Results on HLA-DRB10101 and HLA-DRB10301

Allele	Method	Acc.	A_ROC	A_ROC50	MCC	A_OVER-ROC
HLA-DRB1*0101	SKM (β = 0.04)	0.886	0.912	0.804	0.643	0.088
	SKM (β = 0.085)	0.901	0.904	0.778	0.690	0.096
	LP_top2		0.779			0.221
	TEPITOPE		0.842			0.158
HLA-DRB1*0301	SKM (β = 0.06)	0.763	0.823	0.580	0.531	0.177
	SKM (β = 0.08)	0.757	0.827	0.575	0.525	0.173
	LP_top2		0.721			0.279
	TEPITOPE		0.585			0.415

Results of 5-fold cross-validation with best results shown in bold. Results from LP_top2 and TEPITOPE are taken from [21]. Measurements are same as previously reported, except for AOVER-ROC, which is the area over the ROC curve. AROC = 1.00 is perfect classification, so AOVER-ROC, 1- AROC, can be seen as an error measure.

ISSN: 1471-2105