Table 1

Performance of GSEA and SAM-GS on Test 1. Proportions of randomly generated null gene sets that are identified by each method to be associated with the phenotype (p-value ≤ 0.05) in a mouse-microarray study.

Correlation range from which gene-set members were selected (% of individual genes in the range) [% of individual genes in the range with FDR ≤ 0.01]

Methods

Set Size


10

30

50

100


|r| < 0.01 (.9% of all genes are in the range) [0% with FDR≤0.01]

GSEA

100%

100%

100%

100%


SAM-GS

0%

0%

0%

0%


|r| < 0.02 (2% of all genes are in the range) [0% with FDR≤0.01]

GSEA

100%

100%

100%

100%


SAM-GS

0%

0%

0%

0%


|r| < 0.05 (4% of all genes are in the range) [0% with FDR≤0.01]

GSEA

100%

100%

100%

100%


SAM-GS

0%

0%

0%

0%


|r| < 0.1 (8% of all genes are in the range) [0% with FDR≤0.01]

GSEA

100%

100%

100%

100%


SAM-GS

0%

0%

0%

0%


|r| < .2 (16% of all genes are in the range) [0% with FDR≤0.01]

GSEA

96%

100%

100%

100%


SAM-GS

0%

0%

0%

0%


|r| < .3 (25% of all genes are in the range) [0% with FDR≤.01]

GSEA

13%

100%

100%

100%


SAM-GS

0%

0%

0%

0%


|r| < .4 (36% of all genes are in the range) [0% with FDR≤.01]

GSEA

8%

100%

100%

100%


SAM-GS

0%

0%

0%

0%


|r| < .5 (47% of all genes are in the range) [0% with FDR≤.01]

GSEA

0%

24%

100%

100%


SAM-GS

0%

0%

0%

0%


Dinu et al. BMC Bioinformatics 2007 8:242   doi:10.1186/1471-2105-8-242

Open Data