Protein-RNA interface residue prediction using machine learning: an assessment of the state of the art

BMC Bioinformatics

Table 4 Residue-based evaluation of Sequence Methods on Structure Data

	IDSeq	IDSeq	IDSeq	PSSM Seq	PSSM Seq	PSSM Seq	Smo PSSM	Smo PSSM	Smo PSSM
	NB	LK	RBFK	NB	LK	RBFK	Seq NB	Seq LK	Seq RBFK
RB106Str	0.74 (7.5)	0.73 (9)	0.74 (7.5)	0.76 (5.5)	0.78 (3)	0.81 (1)	0.76 (5.5)	0.77 (4)	0.79 (2)
RB144Str	0.74 (7)	0.73 (9)	0.74 (7)	0.74 (7)	0.79 (2.5)	0.81 (1)	0.75 (5)	0.77 (4)	0.79 (2.5)
RB198Str	0.73 (7)	0.73 (7)	0.73 (7)	0.72 (9)	0.78 (3)	0.80 (1)	0.74 (5)	0.77 (4)	0.79 (2)
Average	0.74 (7.2)	0.73 (8.3)	0.74 (7.2)	0.74 (7.2)	0.78 (2.8)	0.81 (1)	0.75 (5.2)	0.77 (4)	0.79 (2.2)

AUC (averaged over five folds) of sequence methods on structure data using residue-based evaluation. For each dataset, the rank of each classifier is shown in parentheses. Based on average rank, the best sequence method is the SVM classifier that uses the RBF kernel and PSSMSeq as input. (NB - Naïve Bayes, SVM - Support Vector Machine, LK - Linear Kernel, RBFK - Radial Basis Function Kernel).

ISSN: 1471-2105