EMD: an ensemble algorithm for discovering regulatory motifs in DNA sequences

BMC Bioinformatics

Table 1 The prediction accuracy tested on ECRDB62A set.

Algorithm	nPC	nSn	nSp	sPC	sSn	sSp
BP-MD ^a)	0.183	0.215	0.280	0.303	0.428	0.407
AL-BP-MD	0.213	0.262	0.296	0.324	0.456	0.437
AL-BP-MD-MS	0.209	0.255	0.293	0.321	0.423	0.446
AL-BP-MD-ME-MS	0.197	0.238	0.286	0.316	0.438	0.437
AlignACE	0.141	0.218	0.171	0.264	0.351	0.396
BioProspector	0.174	0.205	0.268	0.287	0.415	0.369
MDScan	0.146	0.174	0.223	0.244	0.345	0.349
MEME	0.160	0.260	0.190	0.300 ^d)	0.440	0.430
MotifSampler	0.150	0.180	0.230	0.300	0.320	0.490
RS-AL ^b)	0.139	0.204	0.166	0.229	0.329	0.341
RS-BP	0.150	0.178	0.231	0.262	0.390	0.350
RS-MD	0.107	0.125	0.169	0.170	0.254	0.271
RS-ME	0.133	0.162	0.203	0.213	0.418	0.282
RS-MS	0.127	0.148	0.187	0.235	0.260	0.384
Random ^c)	0.050	0.061	0.083	0.100	0.161	0.146

a) The best algorithm among EMD-X (X = 2~5) are compared with component algorithms, b) the multi-restart algorithms, and c) the random algorithms. The best performances in terms of nPC or sPC among algorithms of a same category are highlighted in bold. d) Both MEME and MotifSampler are highlighted because they have the same performance in terms of sPC.

ISSN: 1471-2105