Table 6.

Classification performance on DPI benchmark with ratio 1:10. Reported are the mean (std) over the 5 best models scored on the test folds. Values in bold indicate the highest metric for each feature type

		DPI-FDA
Feature	Classifier	AUPRC	AUROC	F1
Random	LR	0.129 ± 0.003	0.597 ± 0.001	0.602 ± 0.010
	MLP	0.139 ± 0.006	0.635 ± 0.014	0.594 ± 0.004
	RF	0.192 ± 0.004	0.720 ± 0.005	0.889 ± 0.001
Structural	LR	0.153 ± 0.003	0.690 ± 0.010	0.617 ± 0.010
	MLP	0.244 ± 0.006	0.771 ± 0.003	0.665 ± 0.010
	RF	0.398 ± 0.021	0.871 ± 0.008	0.885 ± 0.016
ComplEx	LR	0.170 ± 0.002	0.706 ± 0.004	0.639 ± 0.004
	MLP	0.339 ± 0.006	0.838 ± 0.004	0.755 ± 0.003
	RF	0.352 ± 0.014	0.849 ± 0.003	0.892 ± 0.008
RotatE	LR	0.180 ± 0.002	0.722 ± 0.005	0.643 ± 0.003
	MLP	0.453 ± 0.008	0.886 ± 0.002	0.794 ± 0.006
	RF	0.404 ± 0.005	0.885 ± 0.006	0.905 ± 0.006
TransE	LR	0.170 ± 0.003	0.691 ± 0.021	0.650 ± 0.017
	MLP	0.350 ± 0.009	0.846 ± 0.005	0.750 ± 0.003
	RF	0.380 ± 0.012	0.868 ± 0.010	0.894 ± 0.015
BioBLP-D	LR	0.177 ± 0.003	0.723 ± 0.003	0.641 ± 0.002
	MLP	0.447 ± 0.008	0.885 ± 0.003	0.792 ± 0.002
	RF	0.397 ± 0.003	0.883 ± 0.004	0.907 ± 0.004
BioBLP-M	LR	0.166 ± 0.002	0.713 ± 0.002	0.634 ± 0.003
	MLP	0.401 ± 0.007	0.865 ± 0.002	0.764 ± 0.002
	RF	0.415 ± 0.003	0.881 ± 0.004	0.895 ± 0.006
BioBLP-P	LR	0.170 ± 0.006	0.683 ± 0.005	0.636 ± 0.005
	MLP	0.343 ± 0.008	0.820 ± 0.005	0.744 ± 0.015
	RF	0.317 ± 0.006	0.832 ± 0.004	0.899 ± 0.002