Biology Reference
In-Depth Information
Metabolic Network Reconstruction
The main sources of information regarding the composition of the metabolic network
of
Pseudomonas putida
KT2440 were various biological databases. Most of the infor-
mation came from the KEGG [35, 81] and Pseudomonas Genome Database (PGD)
[82]. Information regarding
P. putida
contained in these two databases is mainly based
on the published genome annotation of the bacterium [14], so there is a large overlap
between them. Additionally, substantial information was taken from the BRENDA da-
tabase, which catalogs reaction and enzyme information [83]. This all was augmented
with knowledge coming directly from primary research publications. The reconstruc-
tion process was performed in an iterative manner, that is, by adding or removing reac-
tions from the model in between rounds of model testing. First, reaction information
for
P. putida
was collected from KEGG and PGD. Reactions supported by sufficient
evidence and with specific enough functional annotations were incorporated into the
model. For every accepted reaction its reversibility was assessed basing on assign-
ments in KEGG pathways as well as information from BRENDA database. For reac-
tions with inconsistent assignments a decision about reversibility was made basing on
analysis of the reaction as well as its reversibility in other organisms. Hereby, a first
version of the metabolic model was created (iJP815
pre1
).
The next step involved assessing whether the reconstructed metabolic network is
able to produce energy from glucose. This was achieved by running FBA with ATP
production set as the objective function. Subsequently, the ability of the model to
grow
in silico
on glucose was tested. Successful
in silico
growth indicates that ev-
ery chemical compound belonging to the biomass equation can be synthesized from
present sources, using the reactions contained in the model. Since the exact cellular
composition of
P. putida
is not known, the composition of
E. coli
biomass was used as
an approximation. This test was performed by running FBA with production of each
biomass constituent set as the objective. If a compound could not be synthesized, the
gaps in the pathway leading to it were identifi ed manually and a search was performed
for reactions that could fi ll the gaps. If this approach was unsuccessful, gaps were
fi lled with reactions from the
E. coli
model. This yielded the second version of the
reconstruction (iJP815
pre2
).
The third round of reconstruction consisted of two sub-steps. First, the compounds
for which transport proteins exist were identifi ed and appropriate reactions added.
Second, the results of BIOLOG carbon-source utilization experiments were compared
with
in silico
simulations for growth on those compounds. It was assumed that the
ability to grow
in silico
on the particular compound as the sole carbon source approxi-
mates the
in vivo
utilization. For those compounds that did not show
in silico
growth,
a literature search was performed in order to identify possible pathways of utilization.
The results of this search, in the form of reactions and GPRs, were added to the model.
The outcome was the fi nal version of the model (iJP815).
Comparison of Growth Yields with the Continuous Culture Experiments
Growth yields on sources of basic elements (C,N,P,S) were compared with experimen-
tal values obtained by Duetz et al. [37]. The yields of the model were computed using