Table 3

Resource Utilization Comparison.

Sequence Pair

CPU (sec)

RAM (MB)

unconstrained pairSCFG

RD0260 vs RE6781 tRNAs

718

285

M16173 vs X02128 5S

22902

1465

constrained pairSCFG (post 0.95; W = 20)

RD0260 vs RE6781 tRNAs

31

64

M16173 vs X02128 5S

347

248

Dynalign (

<a onClick="popup('http://www.biomedcentral.com/1471-2105/7/400/mathml/M1','MathML',630,470);return false;" target="_blank" href="http://www.biomedcentral.com/1471-2105/7/400/mathml/M1">View MathML</a>

= 15; gap = 0.4)

RD0260 vs RE6781 tRNAs

1761

33

M16173 vs X02128 5S

4928

67

Stemloc (-na 100 -nf 1000)

RD0260 vs RE6781 tRNAs

6

85

M16173 vs X02128 5S

18

193

PMcomp (-p -noLP for RNAfold)

RD0260 vs RE6781 tRNAs

19

52

M16173 vs X02128 5S

33

142

FOLDALIGN (-max_diff 25 -global -score_matrix global. fmat)

RD0260 vs RE6781 tRNAs

25

77

M16173 vs X02128 5S

87

280

The resource requirements for a single representative tRNA and 5S sequence pair for each of the five constrained Sankoff implementations and the unconstrained pairSCFG Sankoff implementation, given as a baseline reference.

The tRNA sequences are of length 78 and 77 respectively. The 5S sequences are of length 117 and 118 respectively. The constrained pairSCFG utilizes the pin selection criteria of posteriors > 0.95 and a protection window of 20. (For these examples the result is 2 pins for the tRNA pair and 3 pins for the 5S pair.) Dynalign is compared using the parameters recommended in [35] of

<a onClick="popup('http://www.biomedcentral.com/1471-2105/7/400/mathml/M1','MathML',630,470);return false;" target="_blank" href="http://www.biomedcentral.com/1471-2105/7/400/mathml/M1">View MathML</a>

= 15 and a gap penalty of 0.4 kcal mol-1 gap-1. Stemloc is compared using the parameters recommended in [31] which computes 100 alignment envelopes (-na 100) and 1000 fold envelopes (-nf 1000). For PMcomp, the reported time includes both calculating the pair probability (via RNAfold with parameters -p -noLP) and the subsequent pmcomp.pl phase. The memory reported is the maximum utilized during either phase. With FOLDALIGN, the parameters from Gardner are utilized (-max_diff 25-global -score_matrix global.fmat). For all programs these are the same parameters utilized to generate the performance reflected in Figure [7]. These benchmarks were conducted on a dual 2.8 GHz P4 machine with 2.5 GB memory running the Linux 2.4 kernel.

Dowell and Eddy BMC Bioinformatics 2006 7:400   doi:10.1186/1471-2105-7-400

Open Data