Evaluating the model

[yunzhang1998]

Hi Author,

Thank you for providing such useful work!

I am currently trying to train a similar model with my own data, but I have encountered some questions. When evaluating the model, were the top-k accuracy and AU-PRC values calculated for each input sequence (1000 bp) and then averaged? And for tissue usage, was the AU-PRC computed only on the splicing sites of each input sequence (may be only one in my data)？

I would greatly appreciate your clarification and look forward to your response.

[SilverNebula]

Hello,
Thank you for your interest in our work.
During the evaluation process, we followed an approach similar to that of SpliceAI and Pangolin by gathering all data entries together to compute the top-k accuracy and AU-PRC values. We did not perform separate calculations or average across individual subsets.

Best regards,
ShenLab Genomics