Skip to main content
. 2023 Nov 6;24(21):16000. doi: 10.3390/ijms242116000

Figure 3.

Figure 3

Determination of the optimal probability threshold cut-off that separates the positive and negative classes using ROC curve by maximizing g-mean of specificity and sensitivity. The figure illustrates the optimal cut-off points determined from training dataset using (a) Ankh embeddings, (b) ESM-2 embeddings, and (c) ProtT5 embeddings.