Deep learning-based classification model for GPR151 activator activity prediction

BMC Bioinformatics

Table 5 Evaluation of neural network models and classifiers on test set performance with processed rdk fingerprint, and the bold marks the best in the table

Feature	Classifier	Accuracy	Precision	Recall	F1	Trend
RDKFP processed	CNN	0.9003	0.7806	0.6296	0.697	\(\uparrow\)
	LSTM	0.8748	0.6759	0.6008	0.6362	\(\uparrow\)
	Bi-LSTM	0.8823	0.724	0.572	0.6391	\(\uparrow\)
	CNN+LR	0.8913	0.7988	0.5391	0.6437	\(\uparrow\)
	CNN+KNN	0.8808	0.7121	0.5802	0.6395	\(\uparrow\)
	CNN+RF	0.8546	0.6992	0.3539	0.4699	\(\uparrow\)
	CNN+DT	0.8493	0.8088	0.2263	0.3537	\(\downarrow\)
	CNN+SVM	0.8913	0.7988	0.5391	0.6437	\(\downarrow\)
	LSTM+LR	0.8748	0.6712	0.6132	0.6409	\(\uparrow\)
	LSTM+KNN	0.8763	0.6857	0.5926	0.6358	\(\uparrow\)
	LSTM+RF	0.8718	0.6765	0.5679	0.6174	\(\uparrow\)
	LSTM+DT	0.8718	0.6765	0.5679	0.6174	\(\uparrow\)
	LSTM+SVM	0.8748	0.6776	0.5967	0.6346	\(\uparrow\)
	Bi-LSTM+LR	0.8808	0.7234	0.5597	0.6311	\(\uparrow\)
	Bi-LSTM+KNN	0.8816	0.7202	0.572	0.6376	\(\uparrow\)
	Bi-LSTM+RF	0.8756	0.6935	0.5679	0.6244	\(\uparrow\)
	Bi-LSTM+DT	0.8711	0.6802	0.5514	0.6091	\(\uparrow\)
	Bi-LSTM+SVM	0.8801	0.7044	0.5885	0.6413	\(\uparrow\)

ISSN: 1471-2105