Ranking metrics in gene set enrichment analysis: do they matter?

BMC Bioinformatics

Table 3 Results of overall sensitivity, false positive rate and average evaluation time for all ranking metrics

Rank metric	\(1-\hat \pi _{0}\)	\( \| \overline {FPR} - \widehat {FPR} \|\)	Average evaluation time [s]
T-test	0.066	8.162	118.363
MWT	0.928	19.634	191.944
\|MWT\|	0.998	3.665	191.944
MSD	0.559	0.008	117.627
S2N	0.926	17.932	115.090
\|S2N\|	0.981	2.542	115.090
WAD	0.992	23.482	112.534
\|WAD\|	0.994	23.971	112.534
Difference	0.985	25.212	111.920
Ratio	0.997	26.429	121.228
log₂(Ratio)	0.824	23.337	120.820
FCROS	0.758	19.228	324.413
SoR	0.756	23.087	264.746
BWS	0.900	2.696	289.215
ReliefF	0.840	6.394	912.471
ReliefF ranked	0.548	2.852	912.471

Overall sensitivity is defined as 1 - estimator from Storey’s method (the higher, the better). Overall false positive rate is defined as an absolute value of the difference between observed and expected false positive rate (the lower, the better)

ISSN: 1471-2105