Table 3.

Comparison of our method (HMM) with RF

Variants	Size	Error rate (%)			False-negative rate (%)			False-positive rate (%)
		RF	HMM	Rel	RF	HMM	Rel	RF	HMM	Rel
AATAAA	5190	20.06	18.59	7.31	19.74	18.54	6.10	20.37	18.65	8.44
ATTAAA	2400	18.42	16.21	12.01	18.68	18.17	2.75	18.15	14.25	21.49
AAAAAG	1250	16.64	9.36	43.75	16.53	11.36	31.28	16.75	7.36	56.06
AAGAAA	1230	11.06	5.45	50.75	11.92	6.02	49.53	10.15	4.88	51.94
TATAAA	880	19.55	15.34	21.53	18.10	19.09	−5.47	20.87	11.59	44.46
AATACA	780	19.36	11.15	42.39	18.13	15.64	13.73	20.49	6.67	67.46
AGTAAA	690	27.83	16.96	39.07	25.24	20.00	20.76	29.92	13.91	53.50
ACTAAA	670	22.09	14.33	35.14	20.69	20.60	0.45	23.36	8.06	65.50
GATAAA	460	20.00	9.57	52.17	21.01	10.43	50.33	18.92	8.70	54.04
CATAAA	410	18.54	9.27	50.01	16.92	14.63	13.51	20.00	3.90	80.49
AATATA	410	24.88	12.68	49.02	24.12	20.49	15.06	25.59	4.88	80.94
AATAGA	370	18.38	5.14	72.06	19.37	6.49	66.51	17.32	3.78	78.15
Average	—	19.19	14.42	25.62	18.83	16.26	14.81	19.48	12.59	35.40

Note: The performance of both RF and HMM is evaluated on the same 5-fold cross-validation. ‘Rel’ denotes the relative improvement of HMM with respect to RF. The lowest value for each criterion of each motif variant is indicated in bold.