The Neural/Immune Gene Ontology: clipping the Gene Ontology for neurological and immunological systems

BMC Bioinformatics

Table 2 An overview of the performance of NIGO, full GO and GO-slim in enrichment analysis using GSEA

		Unique to Subset			Lowest FDR Value
Experiment	Cutoff used	Full GO	NIGO	Generic GO slim	Full GO	NIGO	Generic GO slim
Neural- or immune-related studies
GSE6509	P < 0.05	2 (2)	5	0	0	2	0
GSE6675	FDR < 0.25	0	6	0	0	1	0
GSE6476	FDR < 0.25	1 (1)	11	0	0	4	0
GSE3779	P < 0.05	7 (3)	13	1 (1)	0	1	5
GSE8425	P < 0.05	6 (1)	0	0	3	3	1
GSE6690	P < 0.01	2 (2)	0	0	1	8	8
GSE6136	P < 0.05	3	2	0	0	0	1
GSE8788	P < 0.01	6 (6)	0	0	11	3	1
GSE8788*	P < 0.01	6 (6)	0	0	3	11	1
GSE9659	P < 0.01	6 (6)	0	NS^A	4	28	NS^A
Non-neural- or immune-related studies (negative controls)
GSE7407	P < 0.01	5 (5)	0	0	0	1	2
GSE2259	FDR < 0.25	3 (1)	0	1 (1)	1	2	1
GSE8191	P < 0.05	5 (2)	0	0	0	0	0

The enrichment analysis results of GSEA, providing the full GO, NIGO or GO-slim as the sole source of gene sets, are compared for neural/immune-related (top) and unrelated (bottom) studies. A statistical cutoff was chosen for each analysis such that at least five terms were found to be significantly enriched in any of the analyses. Studies are identified via their GEO accession. The number of terms that passed the statistical cutoff inclusively for each subset are provided ('Unique to Subset'), together with the number of terms that are not included in the NIGO subset out of the total number of terms that passed the statistical cutoff only with the full GO or generic slim (given in parentheses). The number of terms that passed the statistical cutoff with NIGO and at least one additional GO subset (the full GO or generic slim) are also shown for each term ('Lowest FDR Value'), with the number of terms with the lowest FDR values for each subset given in separate columns. *To test the effect of stochasticity, the analysis of this profile is based on 3-fold averaged FDR values from three independent GSEA analyses. ^A Profile GSE9659 involves rat tissue, and was not analyzed with the mouse GO-slim.

ISSN: 1471-2105