Biology Reference
In-Depth Information
its maximum value. In this suboptimal FVA experiment, the
13
C -derived fluxes fall
between FVA intervals for every flux value in the
13
C network. To filter out artifacts,
we re-did all FVA computations using the structure of the network used in the
13
C
-experiment and found no major differences. We also assessed the influence of the
biomass composition on the distribution of internal fluxes and network structure and
found that this was negligible on both accounts. The results show that, in principle,
the bacterium can use all the alternatives described above and that the penalty on the
growth yield is minimal. While this analysis validates the FVA simulation results, the
wide breadth of the intervals (i.e., the mean ratio of interval width to mean interval
value exceeds three), suggests that the (mathematical) under-determination of central
metabolism can be quite high, and indicates that there exist multiple sub-optimal solu-
tions across the network and that is thus difficult to predict exact internal flux and to
“pin-point” a particular solution. These results reflect the essence of CB modeling and
FBA, which provide only a space of possible flux distributions and not exact values.
Therefore, deductions from results of FBA simulations have to be made with great
care. This underscores the notion that constraint-based modeling should be seen more
as navigation framework to probe and explore networks rather than as an exact predic-
tive tool of cellular metabolism.
Gauging the Robustness of the Network
Essentiality of Genes and Reactions
To assess the robustness of the metabolic network to genetic perturbations (e.g.,
knock-out mutations), we carried out an
in silico
analysis of the essentiality of sin-
gle genes and reactions, which enabled us to identify the most fragile nodes of the
iJP815 network. Reaction essentiality simulations were performed by systematically
removing each reaction from the network and by assessing the ability of the model
to produce biomass
in silico
via FBA in minimal medium with a sole carbon source
(glucose and acetate). Gene essentiality was assessed by: (i) identifying for each
gene the operability of the reaction(s) dependent on this gene, (ii) removing from the
network the reactions rendered inoperative by the deletion of that particular gene,
and (iii) determining the ability of the model to produce biomass in the same man-
ner as for the reaction essentiality tests. Additionally, we estimated for both carbon
sources the smallest possible set of reactions able to sustain
in silico
growth, in or-
der to estimate the number of reactions necessary for biomass synthesis in minimal
medium (minimal set). This set encompasses both all reactions that are essential
(including those essential regardless of the medium and those “conditionally essen-
tial”) and the minimal number of non-essential reactions that, together, are able to
provide
in silico
growth (see Figure 5). These conditionally essential reactions can
be used as a reference for identifying sections of metabolism for which alternative
pathways exist. For both glucose and acetate, the minimal sets encompassed ap-
proximately 315 reactions. This estimate is consistent with values obtained for other
bacteria [54].
