Table 1.

Number and fraction (in parentheses) of proteins in each data set for which predicted PWMs significantly match at P ≤ 0.05

Database	Number of proteins	SVM models
		Linear canonical (%)	Linear expanded (%)	Polynomial (%)
JASPAR	57	37 (65)	43 (75)	41 (72)
UniProbe yeast	13	10 (77)	12 (92)	10 (77)
UniProbe mouse	23	14 (61)	14 (61)	16 (70)
FlyFactorSurvey	52	45 (87)	47 (90)	47 (90)
Human	36	25 (69)	29 (81)	28 (78)
Combined	143	107 (75)	118 (83)	114 (80)

The best performing method on each data set is shaded.