Motif Enrichment Analysis: a unified framework and an evaluation on ChIP data

BMC Bioinformatics

Table 1 Specific MEA methods tested in this study.

*Method Name*	*Motif Affinity Function (X* _g )	*Biological Signal (Y* _g )	*Partition Maximization Variants*
Fisher	MC: {0, 1}	1/p	YFP, YCPM, YUPM
mHG	MC: {0, 1, 2}	1/p	YFP, YCPM, YUPM
Ranksum	AMA	1/p	YFP, YCPM, YUPM
Clover	AMA	1/p	YFP, YCPM, YUPM
PASTAA	AMA-like	1/p	X, YCPM
Spearman	AMA	1/p	none
LR	RMA	-log(p)	none

The method names refer to the association function they use: "Fisher" (Fisher Exact Test), "mHG" (multi-hypergeometric test), "Ranksum" (Ranksum or Mann-Whitney U Test), "Clover", "Spearman" (Spearman's rank correlation coefficient) and "LR" (linear regression). For Fisher, the MC motif affinity function is capped at 1 match; for mHG, the MC function is capped at 2 matches. We set t_m= 0.0002 for the MC motif affinity function (i.e., matches have PWM score with p-value less than 0.0002). In all methods except LR, Y_gis defined as 1/p, where p is the p-value of a microarray probe fluorescence signal; for the LR method, Y_gis defined as p.

ISSN: 1471-2105