Table 5. Disease categories clusters analysis between different knowledge graph methods.

Purity and normalized mutual information (NMI) are used to measure clusters’ goodness.

Relation	without t-SNE		with t-SNE
	Purity	NMI	Purity	NMI
Walking RDF/OWL	0.682	0.438	0.507	0.237
TransE	0.540	0.245	0.565	0.254
Poincaré	0.660	0.367	0.569	0.222
Rescal	0.510	0.165	0.434	0.032
SimplE	0.562	0.237	0.556	0.227
R-GCN	0.563	0.226	0.557	0.223

Note:

The highest performing method is indicated in bold while the second highest is underlined.