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

BMC Bioinformatics

Table 3 Performance of motif discovery algorithms on mouse TF ChIP-seq datasets.

	OOPS-vs. OOPS				ZOOPS-vs. ZOOPS				OOPS-vs. ZOOPS-				OOPS-vs. PRIORITY-
TF	W	L	T	p -value	W	L	T	p -value	W	L	T	p -value	W	L	T	p -value
Nanog			×	3.4e-01		×		1.3e-03	×			7.7e-03	×			2.1e-09
Oct4			×	1.0e-01			×	3.4e-01	×			5.8e-07	×			4.5e-14
Sox2			×	1.6e-01		×		1.6e-02			×	4.4e-01	×			1.3e-03
Smad1			×	1.0e-01			×	1.6e-01			×	2.4e-01			×	1.6e-01
E2f1			×	2.4e-01	×			4.5e-05		×		4.5e-05			×	4.4e-01
Tcfcp2l1			×	1.0e-01		×		7.7e-03		×		1.6e-02	×			1.9e-11
Ctcf			×	4.4e-01			×	2.4e-01			×	4.4e-01	×			8.9e-16
Zfx			×	1.0e-01	×			1.3e-03			×	1.6e-01	×			2.2e-10
Stat3	×			3.3e-03			×	4.4e-01			×	6.0e-02			×	1.6e-01
Klf4			×	1.6e-01			×	6.0e-02			×	1.0e-01			×	1.6e-01
Esrrb			×	6.0e-02			×	6.0e-02	×			3.3e-03		×		4.5e-14
c-Myc	×			3.3e-02	×			3.3e-03			×	2.4e-01		×		4.5e-05
n-Myc	×			1.5e-04	×			4.5e-05	×			1.6e-02		×		1.6e-08
Total	3	0	10		4	3	6		4	2	7		6	3	4

The table compares the relative accuracy of pairs of motif discovery algorithms. Relative accuracy is measured by the correlation on held out sets of sequences of the sequence ranks based on ChIP-seq peak scores versus the ranks based on the motif-based AMA score. A check in the "win" or "W" ("loss" or "L") column indicates that the motifs found by the first (second) algorithm had significantly better Spearman rank correlation, as judged by the sign test on the 50 random repeats (p-value < 0.05). A check in the "tie" or "T" column indicates that there was no significant difference. The "Total" line shows the totals using the sign test to judge significance. OOPS, ZOOPS, OOPS- and ZOOPS- refer to MEME with those models and with or without the prior.

ISSN: 1471-2105