Table 2. 10-fold cross-validation on Yam08 dataset.

	BLM-NII	CMF	KBMF2K	NetLapRLS	NRLMF	WNN-GIP	COSINE
AUPR
N. Recept.	0.438	0.488	0.477	0.417	0.545	0.504	0.548
GPCR	0.315	0.365	0.366	0.229	0.364	0.295	0.397
Ion Ch.	0.302	0.286	0.308	0.200	0.344	0.258	0.359
Enzyme	0.253	0.229	0.263	0.123	0.358	0.278	0.346
AVERAGE	0.327	0.342	0.354	0.242	0.403	0.334	0.410
AUC
N. Recept.	0.799	0.818	0.844	0.789	0.900	0.890	0.914
GPCR	0.838	0.857	0.839	0.817	0.895	0.891	0.902
Ion Ch.	0.796	0.743	0.799	0.757	0.813	0.797	0.826
Enzyme	0.813	0.829	0.713	0.786	0.871	0.882	0.888
AVERAGE	0.812	0.812	0.799	0.787	0.870	0.865	0.883

The best results are underlined. Cases where COSINE significantly outperforms the competitor (t-test, p < 0.05) are shown in italic.

The results for other methods were taken from Liu et al.²⁷.