Biology Reference
In-Depth Information
complete but that lack a transport reaction for the initial or final compound. Examples
of pathways lacking a transporter are the degradation of fatty acids and of propanoate.
In addition, there could exist compounds whose production is required only in
certain environmental conditions, for example, under solvent stress, and as such are
not included in generic biomass equation. Pathways synthesizing compounds that are
not included in the biomass equation but that likely are conditionally required include
the synthesis of thiamine, various porphyrins and terpenoids. In this case, reactions
involved exclusively in the production of such compounds would be blocked if no
alternative outlets exist for those pathways. Allowing a non-zero fl ux through these re-
actions would require inclusion into biomass of the conditional biomass constituents,
which in turn would require having various biomass equations for various conditions.
This level of detail, however, is beyond the scope of our initial metabolic reconstruc-
tion and investigation.
The high number of blocked reactions in iJP815 clearly indicates that there are
still vast knowledge defi cits in the model and, thus, in the underlying biochemical and
genomic information. Since a genome-scale metabolic model seeks to incorporate all
current knowledge of an organism's metabolism, these reactions are integral elements
of the metabolic reconstruction and of the modeling scaffold, even if they are not able
to directly participate in steady state fl ux studies. Therefore, the inclusion of these
reactions in the model provides a framework to pin-point knowledge gaps, to include
novel information as it becomes available and to subsequently study their embedding
and function in the metabolic wiring of the cell.
How
P.
putida
Allocates its Resources: Evaluating the Prediction of Internal
Flux Distributions
The assessment performed as described above by means of high-throughput pheno-
typing assays, growth experiments, and continuous cultivations, has shown that the
model is coherent and that it captures the major metabolic features of
P. putida
. We
subsequently used the model to probe the network and to ascertain the distribution of
internal fluxes and properties such as network flexibility and redundancy of particular
reactions. To this end, we predicted the distribution of reaction fluxes throughout the
central pathways of carbon metabolism by FVA, and compared the simulations to
internal fluxes computed from experimentally obtained
13
C data in
P. putida
[49, 50].
Optimal FVA
Genome-scale metabolic networks are, in general, algebraically underdetermined
[41]. As a consequence, the optimal growth rate can often be attained through flux
distributions different than the single optimal solution predicted by FBA simulations.
Therefore we used FVA to explore the network, as this method provides the intervals
inside which the flux can vary without influencing the value of the growth yield (if the
flux of the reaction cannot vary then the range is limited to a single value) [41]. The
results of the simulations are given in Figure 4. As isotopic (
13
C) measurements are
not able to distinguish which glucose uptake route is being used by
P. putida
, all the
fluxes in the
13
C experiment and in the FVA simulations were computed assuming that
