Deep learning-based classification model for GPR151 activator activity prediction

BMC Bioinformatics

Table 7 Comparison of neural network models and classifiers in Mol2vec test set performance, and the bold marks the best in the table

Feature	Classifier	Accuracy	Precision	Recall	F1	Trend
Mol2vec	CNN	0.9085	0.7895	0.679	0.7301	\(\uparrow\)
	LSTM	0.8996	0.7444	0.6831	0.7124	\(\uparrow\)
	Bi-LSTM	0.8988	0.7389	0.6872	0.7122	\(\uparrow\)
	CNN+LR	0.904	0.7626	0.6872	0.7229	\(\uparrow\)
	CNN+KNN	0.8958	0.7524	0.639	0.6904	\(\uparrow\)
	CNN+RF	0.91	0.7971	0.679	0.7333	\(\uparrow\)
	CNN+DT	0.8718	0.6651	0.5967	0.6291	\(\uparrow\)
	CNN+SVM	0.9153	0.7928	0.7243	0.757	\(\uparrow\)
	LSTM+LR	0.8973	0.7431	0.6667	0.7028	\(\uparrow\)
	LSTM+KNN	0.8861	0.7136	0.6255	0.6667	\(\downarrow\)
	LSTM+RF	0.8973	0.7409	0.6708	0.7041	\(\uparrow\)
	LSTM+DT	0.8628	0.65	0.535	0.5869	\(\uparrow\)
	LSTM+SVM	0.8951	0.7191	0.6955	0.7071	\(\uparrow\)
	Bi-LSTM+LR	0.8973	0.7409	0.6708	0.7041	\(\uparrow\)
	Bi-LSTM+KNN	0.8921	0.7302	0.6461	0.6856	-
	Bi-LSTM+RF	0.8973	0.7477	0.6584	0.7002	\(\uparrow\)
	Bi-LSTM+DT	0.8711	0.6621	0.5967	0.6277	\(\uparrow\)
	Bi-LSTM+SVM	0.9003	0.7523	0.6749	0.7115	\(\uparrow\)

ISSN: 1471-2105