Hello, I have retrained the model using your original script and found that the results differ significantly from the original text.

[NewCodeAlg]

Hello, I have retrained the model using your original script and found that the results differ from the original text. May I ask where the problem lies? The R2 calculation result is below 0.3, and the kcat PCC verified at a 10fold is only 0.45.

[zchwang]

Please refer to the parameters shown in Table S6-S8 in the Supporting Information.

[NewCodeAlg]

Thank you for your answer. I have now reduced the learning rate 1e-3 to 1e-5, set the batch size to 8, and set dropout to 0.

[NewCodeAlg]

After adjusting the parameters and retraining catapro, the average PPC, SCC, and R2 of the ten fold cross validation were 0.469, 0.46, and 0.22, respectively. rMSE was 1.34, which is close to the original results of PCC0.497, SCC0.495, and rmse1.329, possibly due to differences in equipment.