TABLE 5.

Performance comparison of RNN and CNN-RNN architectures with or without the attention mechanism.

Species	Models	ACC (%)	F-value (%)	Recall (%)	PRE (%)	MCC
A. thaliana	RNN	85.0	85.1	86.6	83.7	0.701
	RNN_Attention	84.6	85.1	87.1	93.2	0.692
	CNN-RNN	83.9	84.3	86.1	83.9	0.678
	CNN-RNN_Attention	83.9	83.8	84.0	83.7	0.678
C. elegans	RNN	89.3	89.7	93.1	86.5	0.788
	RNN_Attention	89.3	89.5	91.4	87.6	0.788
	CNN-RNN	89.0	89.4	92.3	86.5	0.783
	CNN-RNN_Attention	89.4	89.4	90.3	88.6	0.789
D. melanogaster	RNN	85.9	85.8	87.1	84.8	0.720
	RNN_Attention	87.0	87.4	89.6	85.3	0.742
	CNN-RNN	86.8	87.0	88.7	85.3	0.736
	CNN-RNN_Attention	87.4	87.7	89.6	86.0	0.749

The bold values highlight the best methods of the species (i.e., A. thaliana, C. elegans, and D. melanogaster).