Identification of microRNA precursors based on random forest with network-level representation method of stem-loop structure

Table 1 Comparison with existing methods

Methods	Complete dataset		Training dataset		Testing dataset		Results for testing dataset		Results for Independent dataset
	Pos	Neg	Pos	Neg	Pos	Neg	SE	SP	Plant (Acc)	Virus (Acc)	Total (Acc)
Triplet-SVM	193	1168	163	168	30	1000	0.933	0.881	0.882	0.843	0.877^a
microPred	691	9248	SMOTE Outer-5-fold-CV				0.900	0.973	0.841	0.939	0.853^b
Our method	3928	8897	3000	3000	928	5897	0.873	0.911	0.976	0.913	0.970

Triplet-SVM is a SVM-based method with triplet elements that represent information of pre-miRNA stem-loop structure. There is an extension called MiPred.
microPred combined the new RNAfold-related, Mfold-related, and pair-related features with 29 'global and intrinsic' features introduced in the miPred approach.
^a 178 virus and 1232 plant sequences were used, as samples with multiple loops were filtered out by Triplet-SVM.
^b 196 virus and 1389 (the length less than 300) plant sequences were submitted to microPred web server.

ISSN: 1471-2105