Hi, I was wondering how you calculated the ROC AUC and PR AUC. In the context of protein-ligand binding site prediction only the positives are predicted (predicted pockets). How do you get FN? These would be observed pockets that aren’t predicted. But you don’t have scores for those as they aren’t predicted. And what about the TN? Those would be pockets that don’t exist and are not predicted. How do you deal with this? Many thanks.
Hi, I was wondering how you calculated the ROC AUC and PR AUC. In the context of protein-ligand binding site prediction only the positives are predicted (predicted pockets). How do you get FN? These would be observed pockets that aren’t predicted. But you don’t have scores for those as they aren’t predicted. And what about the TN? Those would be pockets that don’t exist and are not predicted. How do you deal with this? Many thanks.