Integration Strategy Is a Key Step in Network-Based Analysis and Dramatically Affects Network Topological Properties and Inferring Outcomes

Table 10

The topology properties of the molecular complexes found by 11 networks built by integration strategies based on the MCODE clustering algorithm.

	Empirical				Machine learning
	Union	Intersection	2-Vote	3-Vote	Naive Bayes	Bayesian Networks	Logistic Regression	SVM	Random Tree	Random Forest	J48
Proteins	63	5	26	29	55	65	64	61	40	59	48
Interactions	1721	9	169	92	438	1787	1761	1628	702	1497	1028
Diameter	2	2	2	7	5	2	2	2	2	2	2
Degree	54.127	3.6	12.769	5.241	15.927	54.985	54.844	53.377	35.1	50.746	42.833
Density	0.873	0.9	0.511	0.187	0.295	0.859	0.871	0.890	0.9	0.875	0.911
ASP	1.127	1.1	1.489	3.264	2.773	1.141	1.129	1.110	1.1	1.125	1.089
CC	0.903	0.9	0.940	0.816	0.851	0.894	0.899	0.913	0.904	0.895	0.928

Note: Proteins, Interactions, Diameter, Degree, and Density indicate the number of proteins, the number of interactions, network diameter, average degree, and network density, respectively. ASP and CC are the average path length and clustering coefficient, respectively. Bold type indicates the minimum value on an average path length and the maximum value in the other topological properties of empirical and machine learning methods.