Table 2

Execution times for three versions of the OROlac promoter model.

Fully continuous

Fully discrete

Hybrid

R

Mean (s)

Std (s)

Mean (s)

Std (s)

Mean (s)

Std (s)


1

0.26 [1]

0.01

0.36 [1.4]

0.01

0.69 [2.7]

0.01

10

0.27 [1]

0.01

1.15 [4.3]

0.01

6.91 [27]

0.12

100

0.27 [1]

0.01

9.11 [34]

0.28

69.43 [260]

1.13


The mean values represent the number of seconds of CPU time required to run a simulation of 200 cell cycles on an otherwise unloaded 700 MHz PowerPC G4 processor, averaged over a set of 200 different random seeds, identical for each version of the model. "Std" indicates the standard deviation of this same set of 200 runs. Three values of parameter R, a scaling factor giving the relative speed of the reversible reactions in the system, have been used. In each row, the values in square brackets are normalized by the shortest execution time. The fully continuous version runs substantially faster than the other methods, and its execution time does not depend on R. The fully continuous version produces identical histograms to the fully discrete and hybrid methods for GFP (see Fig. 10A), but for the low-number species such as D8 it produces spurious negative values (see Fig. 10B). If accuracy is required for the small-number states, the fully discrete method should be used; note that for this system, the hybrid approach is consistently slower than the fully discrete method.

Adalsteinsson et al. BMC Bioinformatics 2004 5:24   doi:10.1186/1471-2105-5-24

Open Data