Optimizing drug–target interaction prediction based on random walk on heterogeneous networks

Table 1 Shows the recovered fraction values for top 10, 25, 50, 100, 200, 500 and 1,000 ranks with the number of links removed

Number of links removed	AUAC	AUC	BEDROC	EF	AUC (top 10%)
100	0.947	0.991	0.833	9.23	0.867
200	0.938	0.995	0.827	9.100	0.857
300	0.930	0.995	0.818	8.95	0.845
400	0.920	0.991	0.805	8.79	0.830
500	0.916	0.997	0.801	8.71	0.824
600	0.908	0.995	0.789	8.56	0.812
700	0.899	0.981	0.780	8.42	0.802
800	0.885	0.997	0.761	8.20	0.783
900	0.869	0.955	0.741	7.91	0.765
1,000	0.854	0.956	0.715	7.62	0.741

ISSN: 1758-2946