Evolving knowledge graph similarity for supervised learning in complex biomedical domains

Sousa, Rita T.; Silva, Sara; Pesquita, Catia

doi:10.1186/s12859-019-3296-1

BMC Bioinformatics

Table 1 Median of WAFs with alternative methodologies and with evoKGsim for the different PPI datasets

From: Evolving knowledge graph similarity for supervised learning in complex biomedical domains

Dataset	SSM	Single and static combinations					Exhaustive search	Decision	evoKGsim
(#interactions)		BP	CC	MF	Avg	Max	Combinations	Trees
STRING-EC	SimGIC	0.648	0.822	0.670	0.825	0.814	0.825	0.804	0.826
(2245)	Resnik_Max	0.670	0.819	0.641	0.806	0.826	0.817	0.884	0.864
	Resnik_BMA	0.661	0.828	0.642	0.831	0.848	0.832	0.837	0.849
STRING-DM	SimGIC	0.891	0.880	0.791	0.890	0.891	0.927	0.855	0.936
(550)	Resnik_Max	0.910	0.899	0.799	0.927	0.927	0.936	0.917	0.937
	Resnik_BMA	0.928	0.871	0.794	0.936	0.918	0.963	0.927	0.945
BIND-SC	SimGIC	0.849	0.831	0.715	0.854	0.840	0.868	0.830	0.876
(1366)	Resnik_Max	0.883	0.845	0.775	0.904	0.908	0.923	0.890	0.923
	Resnik_BMA	0.864	0.842	0.754	0.901	0.868	0.908	0.872	0.901
DIP/MIPS-SC	SimGIC	0.811	0.776	0.690	0.803	0.779	0.818	0.754	0.825
(13807)	Resnik_Max	0.845	0.798	0.703	0.835	0.838	0.854	0.840	0.849
	Resnik_BMA	0.820	0.788	0.698	0.835	0.822	0.842	0.780	0.843
STRING-SC	SimGIC	0.802	0.764	0.684	0.804	0.780	0.814	0.766	0.817
(30384)	Resnik_Max	0.825	0.788	0.682	0.834	0.826	0.839	0.843	0.843
	Resnik_BMA	0.818	0.784	0.678	0.837	0.817	0.837	0.793	0.838
DIP-HS	SimGIC	0.840	0.746	0.698	0.823	0.768	0.857	0.799	0.861
(2739)	Resnik_Max	0.892	0.829	0.770	0.885	0.867	0.914	0.894	0.894
	Resnik_BMA	0.874	0.773	0.754	0.876	0.811	0.872	0.867	0.881
STRING-HS	SimGIC	0.824	0.769	0.700	0.813	0.786	0.823	0.774	0.830
(6912)	Resnik_Max	0.848	0.763	0.723	0.850	0.811	0.868	0.850	0.867
	Resnik_BMA	0.851	0.792	0.725	0.861	0.815	0.870	0.816	0.876
GRID/HPRD-unbal-HS	SimGIC	0.686	0.652	0.621	0.685	0.664	0.701	0.621	0.694
(31320)	Resnik_Max	0.718	0.674	0.655	0.729	0.702	0.734	0.703	0.734
	Resnik_BMA	0.717	0.678	0.646	0.737	0.697	0.742	0.662	0.742
GRID/HPRD-bal-HS	SimGIC	0.647	0.630	0.618	0.672	0.647	0.674	0.590	0.673
(31349)	Resnik_Max	0.656	0.602	0.590	0.648	0.636	0.664	0.636	0.654
	Resnik_BMA	0.652	0.640	0.597	0.673	0.659	0.674	0.604	0.677

In bold, the best result for each dataset-SSM pair. The median WAF achieved for each baseline is underlined when evoKGsim significantly outperforms the baseline (using α=0.01)

Back to article page

ISSN: 1471-2105

Contact us

General enquiries: journalsubmissions@springernature.com