Table 2.

Area under the ROC curve estimates for PPI/DTI datasets using 10-fold cross-validation

Method/dataset	(Hyper)edge classification					Link prediction
	PC	ID	AI	BC	MP	EC	SP	RN	MM	CE	AT	EZ	IC	GR	NR
	Without domain information, $|\Sigma |=1$
Section hypergraphlet kernel (τ = 0)	0.647	0.555	0.529	0.546	0.585	0.724	0.639	0.671	0.628	0.762	0.578	0.537	0.542	0.552	0.783
Subhypergraphlet kernel (τ = 0)	0.838	0.633	0.675	0.830	0.787	0.692	0.691	0.651	0.612	0.873	0.681	0.788	0.711	0.579	0.732
Section hypergraphlet kernel (τ = 1)	0.649	0.551	0.521	0.570	0.599	0.708	0.641	0.678	0.628	0.772	0.581	0.535	0.550	0.542	0.791
Subhypergraphlet kernel (τ = 1)	0.834	0.632	0.672	0.827	0.767	0.681	0.686	0.649	0.611	0.871	0.683	0.784	0.702	0.580	0.736
L3 framework	–	–	–	–	–	0.717	0.577	0.634	0.573	0.511	0.501	0.629	0.642	0.596	0.505
Preferential attachment	–	–	–	–	–	0.814	0.492	0.552	0.498	0.660	0.449	0.534	0.561	0.545	0.497
	With domain information, $|\Sigma |=\{4,8,16\}$
	$\Sigma ={\Sigma }_{\text{GO}}$	$\Sigma ={\Sigma }_{\text{GO}}\cup {\Sigma }_{\text{SS}}$		$\Sigma ={\Sigma }_{\text{GO}}$		$\Sigma ={\Sigma }_{\text{GO}}$						$\Sigma ={\Sigma }_{\text{GO}}\cup {\Sigma }_{\text{SS}}$
Random hyperwalk	0.741	0.626	0.807	0.792	0.936	0.815	0.610	0.599	0.588	0.597	0.608	0.523	0.543	0.534	0.548
Cumulative random hyperwalk	0.711	0.837	0.822	0.800	0.937	0.848	0.681	0.647	0.670	0.569	0.637	0.844	0.617	0.550	0.518
Pairwise spectrum kernel ($k=\{3,4,5\}$)	0.780	–	–	–	–	0.816	0.584	0.716	0.689	0.754	0.629	–	–	–	–
Section hypergraphlet kernel (τ = 0)	0.651	0.634	0.681	0.663	0.784	0.861	0.677	0.698	0.655	0.804	0.559	0.633	0.583	0.625	0.775
Subhypergraphlet kernel (τ = 0)	0.940	0.936	0.918	0.935	0.952	0.796	0.740	0.676	0.708	0.846	0.678	0.936	0.879	0.724	0.757
Section hypergraphlet kernel (τ = 1)	0.650	0.635	0.693	0.674	0.795	0.875	0.697	0.706	0.667	0.786	0.574	0.637	0.585	0.619	0.777
Subhypergraphlet kernel (τ = 1)	0.940	0.939	0.922	0.930	0.952	0.812	0.750	0.719	0.736	0.860	0.728	0.944	0.884	0.747	0.764

Note: The highest performance for each dataset is shown in boldface.