Table 3.

Results of MOST evaluation on Mixed signals dataset

	No. of 6 bp words	TP	FN	Sensitivity
Occobs/Occexp
1.0	1164	41	4	0.91
1.1	1152	41	4	0.91
1.3	929	37	8	0.82
1.5	710	36	9	0.80
1.7	560	35	10	0.78
2.0	382	29	16	0.64
2.2	334	27	18	0.60
2.5	221	27	18	0.60
3.0	109	4	41	0.09
Seqobs/Seqexp
1.0	1173	34	11	0.76
1.1	999	34	11	0.76
1.3	773	34	11	0.76
1.5	579	34	11	0.76
1.7	469	28	17	0.62
2.0	246	21	24	0.47
2.2	293	18	27	0.40
2.5	102	4	41	0.09
3.0	48	1	44	0.02

Experiments on MOST first step: identification of surprising words. The sensitivity of MOST first phase, carried out with different overrepresentation measures, was evaluated. All the 6 bp sequences representing known binding sites, or all of the substrings of binding sites whose length exceeded six, were searched in the list of 6 nt strings extracted as overrepresented, according to different measures and different thresholds (first column). The sensitivity has been calculated as the number of known sites represented in the list over the total number of known sites [sensitivity = TP/(TP+FN); TP, true positives; FN, false negatives].