IPMiner: hidden ncRNA-protein interaction sequential pattern mining with stacked autoencoder for accurate computational prediction

Table 1 Performance comparison between different layer architectures on RPI488

Architecture	Method	Accuracy	Sensitivity	Specificity	Precision	MCC	AUC
Sep-256-128-64	IPMiner	0.891	0.939	0.831	0.945	0.784	0.914
	SDA-RF	0.880	0.922	0.827	0.928	0.762	0.904
	SDA-FT-RF	0.881	0.916	0.831	0.926	0.762	0.909
Con-256-128-128	IPMiner	0.872	0.893	0.843	0.894	0.743	0.903
	SDA-RF	0.884	0.924	0.831	0.934	0.770	0.911
	SDA-FT-RF	0.864	0.885	0.836	0.887	0.727	0.898
Raw input	RPISeq-RF	0.880	0.926	0.822	0.932	0.762	0.903
Raw input	lncPro	0.870	0.900	0.827	0.910	0.740	0.901

Raw input is concatenation of 3-mer frequency features of protein and 4-mer frequency features of RNA
The boldface indicates this measure performance is the best among the compared methods for individual dataset

ISSN: 1471-2164