Biology Reference
In-Depth Information
TABLE 22.1
Software for Network Construction and Network Modeling
Name
Source
Implementation environment
BNM
ΒΌ
Standalone
BooleanNet
code.google.com/p/booleannet/
Python
BoolNet
cran.r-project.org/web/packages/BoolNet/
R/Bioconductor
CellNetAnalyzer
Matlab
CellNetOptimizer
sites.google.com/site/saezrodriguez/software/cellnetoptimizer
Matlab; R/Bioconductor
DDlab
Standalone
DVD v1
dvd.vbi.vt.edu/cgi-bin/git/dvd.pl
Web-based
Learnboo
Matlab
NetBuilder
strc.herts.ac.uk/bio/maria/NetBuilder/
Standalone
Odefy
Matlab
PBN toolbox
code.google.com/p/pbn-matlab-toolbox/
Matlab
RBN toolbox
Matlab
RBNLab
sourceforge.net/projects/rbn/
Java
network kernel, however, is a part of a larger (extended)
regulatory network, which includes a broader range of
regulators (such as Dax1, Rex1, Tbx3, Esrrb, Klf5)
[34,50,105,106]
(shown in
Figure 22.2
).
Analyses of the published studies and available
networks suggest the presence of at least two types of
network domain linked to the core pluripotency circuit: (1)
upstream 'sensory' network elements related to interpre-
tation of extrinsic signals and communication with the core
pluripotency circuit; and (2) downstream 'executive'
elements that interpret states of the core pluripotency
network and transmit this information to epigenetic modi-
fiers and downstream lineage commitment genes. Inter-
estingly, 'sensory' and 'executive' domains as well as the
core networks are represented by several transcription
factors with redundant properties, such as Smads or Klfs.
This network shows an emerging layout for partitioning the
large network into smaller units, kernels and plug-ins,
which can be utilized for the construction of independent
biochemical reaction network (BRN) and models. Conse-
quently, such robust elements with known input/output
characteristics may be integrated into a large model
explaining the properties of the entire pluripotency
network, in a way similar to the assembly of computer parts
on a motherboard (see
Figure 22.4
). Straightforward
construction of BRN networks from integrated data (such
as co-expression networks) is difficult, as BRNs involve
multiple unknown parameters corresponding to binding
constants, synthesis and degradation rates. Accounting for
all
ambiguity, so the preliminary Boolean network analysis
and partition into independent kernels seems to be
a necessary step. In addition, kernels may correspond to
different molecular mechanisms, such as transcriptional
regulation or signal transduction, and combining such
diverse chemical reactions with different kinds of chemical
constants is yet another problem.
A future integrated model, created from the small robust
kernels, should predict diverse outcomes, such as alterna-
tive differentiation to endoderm/trophectoderm lineages in
response to Gata6/Cdx2 balance (see
Figure 22.3
D) or
alternative differentiation to mesendoderm/neuroectoderm
fate in response to Oct4/Sox2 balance
[22,56,57]
or
suppression of commitment to a neural fate in response to
BMP signaling
[6,107,108]
.
Role of Transcription Regulatory Signals
and Transcriptional Gene Networks
Sometimes, molecular mechanisms of gene interactions are
difficult to capture by the Boolean network models
described above. For instance, high or low concentrations of
Oct4 trigger differentiation, whereas moderate Oct4 levels
promote pluripotency
[109,110]
. This rich response by Oct4
target genes resembles the concentration-dependent
response of Hunchback (Hb) target genes in Drosophila
development
[66,111
113]
. Strikingly, in the fly develop-
ment, graded responses to upstream regulators are the key
feature, important for embryo development and morpho-
logical specification
[114,115]
. Boolean models are not
e
these parameters typically results in high levels of
