Fig. 3
From: A general model for predicting enzyme functions based on enzymatic reactions
a The figure illustrates a comparison between BERT classification results with and without pre-training on USPTO datasets. It is evident that the pre-trained BERT model consistently outperforms the non-pre-trained model across all evaluation metrics; b The figure below displays the prediction results of BEC-Pred for 7 categories in the test set. Different colors represent the three evaluation indicators. Remarkably, the bc-pred model achieved the highest accuracy for categories 1, 2, 4 and 6 among all seven categories. Moreover, the model demonstrated moderate accuracy for categories 3 and 5, while the category 7 exhibited relatively lower accuracy