Candidate gene prioritization by network analysis of differential expression using machine learning approaches

Nitsch, Daniela; Gonçalves, Joana P; Ojeda, Fabian; de Moor, Bart; Moreau, Yves

doi:10.1186/1471-2105-11-460

BMC Bioinformatics

Table 1 Overview of prioritization results.

From: Candidate gene prioritization by network analysis of differential expression using machine learning approaches

			STRING v. 7.1			STRING v. 8.2
			top 10	top 20	AUC	top 10	top 20	AUC
	Standard genetic procedure: Simple Expression Ranking		20	25	0.801	20	25	0.801
	Direct Neighborhood Ranking	log2 ratio	27	31	0.859	12	23	0.747
	ε >0.15, a = 0.5	sign. log2 ratio	28	31	0.880	12	24	0.760
		test statistic	29	30	0.856	13	22	0.738
	Kernel Ridge Regression Ranking	log2 ratio	23	29	0.809	14	26	0.759
	λ = 1, nn = 50, K_LED , ^{α = 0.5}	sign. log2 ratio	27	32	0.868	17	25	0.817
RRMA		test statistic	20	26	0.771	15	20	0.691
	Heat Kernel Diffusion Ranking	log2 ratio	32	34	0.900	32	33	0.913
	all expression values for p₀ , m = 2, α = 0.5	sign. log2 ratio	31	34	0.910	31	35	0.923
		test statistic	32	34	0.901	32	34	0.911
	Arnoldi Diffusion Ranking	log2 ratio	27	31	0.857	27	29	0.851
	m = 2, α = 0.5	sign. log2 ratio	28	31	0.885	28	31	0.873
		test statistic	28	30	0.855	28	29	0.844
	Standard genetic procedure: Simple Expression Ranking		18	24	0.777	18	24	0.777
	Direct Neighborhood Ranking	log2 ratio	27	31	0.874	12	23	0.761
	ε >0.15, a = 0.5	sign. log2 ratio	25	28	0.855	11	22	0.736
		test statistic	27	31	0.863	12	24	0.750
	Kernel Ridge Regression Ranking	log2 ratio	21	28	0.769	17	24	0.756
	λ = 1, nn = 50, K_LED , ^{α = 0.5}	sign. log2 ratio	27	31	0.835	20	26	0.796
GCRMA		test statistic	19	22	0.745	16	23	0.744
	Heat Kernel Diffusion Ranking	log2 ratio	31	33	0.905	32	34	0.914
	all expression values for p₀ , m = 2, α = 0.5	sign. log2 ratio	28	33	0.889	29	34	0.907
		test statistic	32	33	0.895	32	35	0.913
	Arnoldi Diffusion Ranking	log2 ratio	27	32	0.875	26	31	0.874
	m = 2, α = 0.5	sign. log2 ratio	25	29	0.860	25	28	0.852
		test statistic	27	31	0.865	28	31	0.862
	Standard genetic procedure: Simple Expression Ranking		24	28	0.837	24	28	0.837
	Direct Neighborhood Ranking	log2 ratio	23	27	0.846	9	20	0.743
	ε > 0.15, a = 0.5	sign. log2 ratio	25	28	0.844	11	22	0.729
		test statistic	27	30	0.855	12	24	0.745
	Kernel Ridge Regression Ranking	log2 ratio	18	24	0.736	17	24	0.766
	λ = 1, nn = 50, K_LED , ^{α = 0.5}	sign. log2 ratio	23	29	0.834	17	22	0.755
MAS5		test statistic	13	18	0.790	16	24	0.790
	Heat Kernel Diffusion Ranking	log2 ratio	26	32	0.877	28	34	0.890
	all expression values for p₀ , m = 2, α = 0.5	sign. log2 ratio	26	31	0.877	26	34	0.899
		test statistic	32	32	0.890	31	34	0.904
	Arnoldi Diffusion Ranking	log2 ratio	25	27	0.849	24	27	0.851
	m = 2, α = 0.5	sign. log2 ratio	25	29	0.853	25	27	0.847
		test statistic	26	30	0.858	27	29	0.850

The results are based on optimized parameter settings for all presented strategies using different STRING networks. Note that the results of the Simple Expression Ranking do not depend on the network because they are only using differential expression levels to rank the candidate genes.

Back to article page

ISSN: 1471-2105

Contact us

General enquiries: journalsubmissions@springernature.com