Table 1 Open in new tab The performance...

Table 1

The performance of iMFP-LG and its variants on the MFBP dataset

Model	Precision ↑	Coverage ↑	Accuracy ↑	Absolute true ↑	Absolute false ↓
iMFP-LG	0.797	0.803	0.796	0.788	0.078
w/o ad^a	0.777	0.785	0.776	0.767	0.082
w/o GAT^b	0.785	0.791	0.784	0.776	0.080
w/o pretrain^c	0.754	0.769	0.752	0.733	0.095

Model	Precision ↑	Coverage ↑	Accuracy ↑	Absolute true ↑	Absolute false ↓
iMFP-LG	0.797	0.803	0.796	0.788	0.078
w/o ad^a	0.777	0.785	0.776	0.767	0.082
w/o GAT^b	0.785	0.791	0.784	0.776	0.080
w/o pretrain^c	0.754	0.769	0.752	0.733	0.095

Note: The highest values are highlighted in bold. ↑ means that a larger value is better on this metric; ↓ means that a smaller value is better on this metric; ^a w/o ad is a variant in which the adversarial training is not used during training process; ^b w/o GAT is a variant without GAT; ^c w/o pretrain is a variant in which the protein language model is re-initialized randomly instead of using pre-trained weights. GAT, graph attention network; MFBP, multi-functional bioactive peptide.

This Feature Is Available To Subscribers Only