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

BMC Bioinformatics

Table 2 Results from varying substitution matrices

AAindex reference	Substitution matrix	Best β	Acc.	A_ROC	A_ROC50	MCC
HENS920102	BLOSUM62. Matrix based on possible pair-wise substitutions from aligned segments of polypeptides [34]	0.05	0.8049	0.8708	0.532	0.543
BLAJ010101	Matrix built from structural superposition data for identifying potential remote homologues [52]	0.027	0.8207	0.8752	0.540	0.571
DOSZ010104	SM_THREADER_NORM. Amino acid similarity matrices based on force fields (Normalised version) [36]	0.045	0.8217	0.8753	0.548	0.572

Evaluating performance for using different scoring matrices using 10-fold CV. The test measurements are same as in previous experiment. Best values are shown in bold.

ISSN: 1471-2105