Table 9. Performance comparison between this method and other BERT-based methods.

Comparison the performance of DNABERT-CBL with other BERT-based models on key metrics such as accuracy (Acc), F1-score, MCC, precision, sensitivity (Sn), and AUC. DNABERT-2_CA_BL outperforms other models like BERT-Promoter and msBERT-Promoter across multiple metrics, achieving high accuracy, F1, and AUC scores. The maximum value of each column is in bold.

Model	Acc	F1	MCC	Precision	Sn	AUC
DNABERT-2_C M _BL	0.858	0.857	0.727	0.869	0.858	0.924
DNABERT-2_C A _BL	0.867	0.866	0.742	0.875	0.867	0.903
BERT-Promoter	0.848	0.846	–	0.864	0.829	0.918
msBERT-Promoter	0.836	0.833	0.696	0.861	0.836	0.939
TSSNote-CyaPromBERT	0.822	0.822	–	0.824	0.820	0.899