The value of position-specific priors in motif discovery using MEME

BMC Bioinformatics

Table 2 Performance of motif discovery algorithms on yeast TF ChIP-chip datasets.

*Algorithm*	*Description*	*Successes*	*Proportion of Successes*
PhyloCon	local alignment of conserved regions	19	12%
PhyME	alignment-based; uses EM	21	13%
MEME_c	MEME run with non-conserved bases masked	49	31%
PhyloGibbs	similar to PhyME but uses Gibbs sampling	54	35%
Kellis et al.	alignment-based	56	36%
Converge	alignment-based; uses EM	66	42%
PRIORITY-	Gibbs sampler with conservation-based priors	69	44%
PRIORITY-	Gibbs sampler with discriminative conservation-based priors	76	49%
MEME: OOPS	MEME with OOPS model	36	23%
MEME: ZOOPS	MEME with ZOOPS model	39	25%
MEME: OOPS-	MEME with OOPS model and priors	73	47%
MEME: ZOOPS-	MEME with ZOOPS model and priors	81	52%
PRIORITY-	Gibbs sampler with discriminative conservation-based priors	69 (3)	44%

The table shows the number motifs (out of 156) successfully discovered by the named algorithms. The results in the top half of the table are taken from Gordân et al. [8]. Results in the bottom half are for new experiments performed by us. Each algorithm is allowed to report one motif, and success is declared if the scaled Euclidean distance to the known PSPM is <0.25. Proportions (out of 156) successes are rounded to the nearest integral percent.

ISSN: 1471-2105