Table 2.

Computational complexity and running times of the tested algorithms

	Computational complexity	Running time (s)
		BRAliBase	RNAspa	RNA STRAND	RRS
needle	O(n∧2)	1.2	19.8	32.6	105.9
MBR+BONUS	O(n∧2)	0.4	4.1	1.7	12.6
gardenia	O(n∧4)	1.2	36.4	20.8a	327.1
RNAStrAT	O(n∧4)	2.3	630.2	1498.3	11403.4
LocARNA	O(n∧2(n∧2+m∧2))	2.2	166.5	64.5	982.3
RNAdistance	O(n∧3)	1.2	19.9	32.7	109.7
RNAforester	O(|F1|*|F2|*deg(F1)*deg(F2)*(deg(F1)+deg(F2))b	3.3	1460	2654	2 days

For each data set, the running time (computed on a Intel® Core™ i7–2600K CPU @3.40 GHz with 16GB RAM) is reported in seconds employed to process the whole data set. The modified Needleman–Wunsch algorithm that can take as an input BEAR strings and can use the MBR as a substitution matrix was implemented in Java.

^aAn output alignment was produced for only 70% of the total input.

^b|Fi| is the number of nodes in the forest Fi; deg(Fi) is the degree of Fi.