Table 4.

Note the improvement in prediction accuracy on the supposedly more difficult and longer E. coli RNase P data-set. This shows that MFE methods are less sensitive to folding errors on longer data-sets but are also less likely to resolve the entire structure. There is little difference in algorithm accuracy for each of the methods explored here. Each employs the same energy parameters so differences are due to slightly different implementations.

	E. coli RNase P: Single Sequence Methods
Algorithm	number of bps in reference	number of bps in prediction	True Positives (% sensitivity)	False Positives (% selectivity)	Correlation (%)
RNAfold	110	116	69 (62.7)	46 (60.0)	0.612 (61.4)
Mfold (1)	110	118	67 (60.9)	49 (57.8)	0.591 (59.3)
Mfold (2)	110	114	67 (60.9)	46 (59.3)	0.599 (60.1)
Mfold (3)	110	118	76 (69.1)	37 (67.3)	0.680 (68.2)
Sfold (1)	110	116	73 (66.4)	42 (63.5)	0.647 (64.9)
Sfold (2)	110	119	86 (78.2)	28 (75.4)	0.767 (76.8)
Sfold (3)	110	117	61 (55.5)	55 (52.6)	0.538 (54.0)