Table 3.

The F₁ score, MCC and AUC of subcellular location prediction generated by model BLSTM, BLSTM + ConvNet1, ConvNet2 and BLSTM + ConvNet1 + ConvNet2

	D3106			D4802
	F1	MCC	AUC	F1	MCC	AUC
BLSTM	0.7473	0.6001	0.9242	0.7419	0.5705	0.9121
BLSTM + ConvNet1	0.7775	0.6419	0.9255	0.7801	0.6284	0.9327
ConvNet2	0.6475	0.4819	0.8785	0.6696	0.4259	0.9297
BLSTM + ConvNet1 + ConvNet2	0.7843	0.6410	0.9458	0.7842	0.6411	0.9434

The four models were tested on datasets D3106 and D4802. On dataset D3106, the highest F₁ score and AUC are achieved by the model BLSTM + ConvNet1 + ConvNet2, while the model BLSTM + ConvNet1 has the highest MCC. On dataset D4802, the model BLSTM + ConvNet1 + ConvNet2 was the best among the four models