Biology Reference
In-Depth Information
FIGURE 10.8
A four-node signal transduction network in which A is activated by Input, B is inhibited by Input, the activation of the Output
requires the presence of both A and B. (A) The network representation; (B) the list of Boolean rules.
results as possible. In a noteworthy example, Samaga et al.
performed a comparison between simulation results
generated by a Boolean logic model based on the literature
and data collected from cells on ErbB receptor phospho-
protein signaling
[54]
. They came to a set of 11 hypotheses
regarding ErbB signaling resulting from the discrepancies
between simulation results and experimental observations,
among which five were supported by the literature, five led
to further modifications of the model, and one implied the
absence of specificity expected from a small molecular
inhibitor.
Output of the system will not be affected, owing to the OR
relation in the Boolean rule of node C. However, if the edge
from Input to C becomes non-functional, a new oscillatory
behavior of the system is exhibited. It is also possible that
existing attractor basins get altered, or completely elimi-
nated, and new ones arise from such perturbations of the
system.
Validated models can be used to predict the outcomes of
'what if' scenarios
e
cases that have not yet been studied
experimentally
e
and can generate testable predictions and
significant insights. For example, Mendoza formulated
a logic-based model of interactions among cytokines and
transcription factors in helper T (Th) cells
[28]
. Dynamic
simulation under all combinations of initial node states
revealed four steady states: one corresponding to na¨ve Th
cells, one corresponding to Th2 cells, and two corre-
sponding to Th1 cells. The two Th1 cell attractors indicated
two Th1 cell subpopulations with different levels of IFN-
g
secretion, but the level of the IFN-
g
receptor was the same
in both attractors, a result supported by the literature.
Mendoza studied in detail how node perturbations
(knockout or over-expression) change the differentiation
fate of Th cells. Several of the model results were supported
by the experimental literature data, but numerous others are
novel predictions
[28]
.
We next take a closer look at Boolean models applied in
two different contexts: survival signaling in T cells in T-
LGL leukemia
[12]
, and interactions between pathogens
and a mammalian immune system
[11]
. The second
example also demonstrates that Boolean modeling is
a general approach that can be applied at different levels of
biological organization, from the molecular to the pop-
ulation level
[60]
.
Robustness against Disruptions and Useful
Implications
An important additional assessment is whether the model is
robust in terms of changes in interactions or Boolean
transfer functions. Models that are extremely fragile to such
changes may not be a good representation to biological
systems, as the real systems exhibit substantial robustness
to changes in concentrations, reaction rates or even muta-
tions
[55
e
59]
. As currently comprehensive models of
signal transduction systems are rare, the model should show
reasonable robustness to changes in the network structure
to instill confidence that its results will still stand after new
components or interactions are discovered. The ability of
the model to maintain the wild-type response under small
topological perturbations can be tested by adding or
deleting a randomly selected node or edge, rewiring edges
in the network randomly (for example changing any pair of
parallel edges to cross-edges) or making an inhibitory
interaction into activation or vice versa. For example, if in
Figure 10.1
node B is knocked out (constitutively OFF), the
