Table 4.

Sensitivities of the Dynalign/LIBSVM classifier, RNAz, and QRNA broken down by percent identity.

		% sensitivity					% specificity
% identity range		N	% of real set	Dynalign (P > 0.819)	RNAz (P > 0.789)	QRNA	N	% of negative set	Dynalign (P <= 0.819)	RNAz (P <= 0.789)	QRNA
[0	10)	0	0	N/A	N/A	N/A	0	0	N/A	N/A	N/A
[10	20)	0	0	N/A	N/A	N/A	0	0	N/A	N/A	N/A
[20	30)	3	0.0033	100.0	0.0	66.6667	49	0.0271	97.9592	100.0	95.9184
[30	40)	1337	1.4767	71.4286	34.6298	48.6163	9037	4.9907	99.1922	97.0233	97.8975
[40	50)	22328	24.6612	63.6465	63.1539	46.5559	85008	46.9455	99.3612	98.8542	99.2789
[50	60)	42733	47.1984	73.0606	75.6488	56.2469	55602	30.7061	99.2320	99.5018	99.0306
[60	70)	17346	19.1586	88.4930	92.2461	86.0775	23952	13.2274	99.1566	99.3028	98.8143
[70	80)	4061	4.4854	94.5826	94.5087	94.1394	4672	2.5801	97.5813	98.4375	97.9238
[80	90)	2035	2.2477	95.5774	91.9410	93.9066	2062	1.1387	90.8341	98.1086	96.6052
[90	100)	654	0.7223	98.4709	72.6300	59.7859	654	0.3612	60.5505	96.7890	95.2599
[100 ]		42	0.0464	92.8571	61.9048	11.9048	42	0.0232	61.9048	90.4762	95.2381
totals		90539	100.0	75.3366	81.2854	62.0108	181078	100.0	98.9938	98.9927	98.9905

A comparison of sensitivities of the three ncRNA classification/detection programs within each percent identity range. N is the number of sequences within each range. Probability cutoffs for RNAz and the Dynalign/LIBSVM classifier were selected such that overall specificities for the entire test set match the specificity of QRNA as closely as possible.