Table 2.

True positive rates at 1% false positive rate

TF	Array	8-mer	MatrixREDUCE	MDScan	PREGO	RankMotif++	Seed and Wobble	kmerHMM
Cbf1	#1	0.515	0.39	0.231	0.362	0.493	0.383	0.515
	#2	0.459	0.348	0.202	0.336	0.424	0.284	0.462
Ceh-22	#1	0.37	0.26	0.316	0.225	0.427	0.254	0.380
	#2	0.257	0.226	0.293	0.2	0.332	0.251	0.317
Oct-1	#1	0.474	0.365	0.274	0.339	0.315	0.239	0.440
	#2	0.382	0.31	0.213	0.274	0.24	0.202	0.314
Rap1	#1	0.257	0.197	0.213	0.197	0.247	0.226	0.274
	#2	0.277	0.171	0.32	0.179	0.325	0.28	0.243
Zif268	#1	0.449	0.332	0.335	0.328	0.33	0.336	0.439
	#2	0.431	0.297	0.314	0.301	0.389	0.313	0.413

Given the binding preferences of each method on different data sets, their sensitivities (true-positive rates) were computed at the 99% specificity level (false-positive rate). The performance values of all the methods except kmerHMM are adopted from Table 1 on the RankMotif++ manuscript (57). The highest values (except the 8-mer gold standard) are highlighted in bold. The 8-mer gold standard is the method in which the maximum of the median binding intensities of the 8-mers on a testing probe (60 bp) is used as the predicted binding preference of the testing probe (60 bp).