An effective and effecient peptide binding prediction approach for a broad set of HLA-DR molecules based on ordered weighted averaging of binding pocket profiles

Table 3 The performance of OWA-PSSM, TEPITOPE and TEPITOPEpan on the SMM-align dataset in terms of AUC.

Allele	# peptides	TEPITOPE	TEPITOPEpan	OWA-PSSM
DRB1*0101	1203	0.647	0.648	0.645
DRB1*0301	474	0.733	0.739	0.731
DRB1*0401	457	0.754	0.770	0.756
DRB1*0404	168	0.829	0.832	0.830
DRB1*0405	171	0.789	0.785	0.789
DRB1*0701	310	0.768	0.768	0.771
DRB1*0802	174	0.769	0.774	0.784
DRB1*0901	117		0.686	0.731
DRB1*1101	359	0.709	0.700	0.715
DRB1*1302	179	0.721	0.728	0.727
DRB1*1501	365	0.725	0.727	0.731
DRB3*0101	102		0.724	0.869
DRB4*0101	181		0.722	0.729
DRB5*0101	343	0.654	0.652	0.646
Average			0.732	0.747
Average I		0.736	0.738	0.739
Average II			0.711	0.776

"Average" is the average over 14 alleles. "Average I" is the average over 11 alleles predictable by TEPITOPE. "Average II" is the average over 3 alleles not predictable by TEPITOPE. We obtain the PSSMs of TEPITOPE and TEPITOPEpan from their public servers ProPred (http://www.imtech.res.in/raghava/propred/page4.html) and TEPITOPEpan (http://www.biokdd.fudan.edu.cn/Service/TEPITOPEpan/TEPITOPEpan.html), respectively.

ISSN: 1477-5956