Biomedical Engineering Reference
In-Depth Information
Fig. 3.2
p
53 pathways
dna_dsb
ATM
p53
Wip1
Mdm2
3.4
Experimental Results
3.4.1
Model Implementation
We evaluate the SAT-based method for determining the BN on two GRNs, one
synthetic (randomly generated) and one real (
p53 network
network [
4
], [
21
]). We
first investigate the senstivity of our method regarding the number of available gene
expression observations, and then we demonstrate our method on attractor data from
the
p
53 network.
The
p53
network is well-studied in genomics and medicine, due to the involve-
ment of
p53
gene in many human cancers.
p
53 is a tumor suppressor gene and is a
transcription factor for many downstream genes involved in controlling cell cycle,
repairing DNA damage, and inducing apoptosis (cell death) for example. The main
pathways for
p
53 [
22
] involve DNA damage in the form of breaks in the DNA strand,
as shown in Fig.
3.2
In the figure, forward arrows represents activation, while "ar-
rows" with a perpendicular line represents repression. The presence of the external
signal dna_dsb (DNA strand break damage) activates
ATM
, which in turn represses
Mdm
2, allowing for activation of
p
53. The expression of
p
53 blocks replication
of DNA (a necessary response when DNA is damaged). From these pathways, [
4
]
obtained the corresponding Boolean functions.
In our experiments, the function of each gene in both networks is known, but
hidden from our algorithm. We extract both the predictor set and gene expression
observations to test our algorithm with. The regulating logic functions of the synthetic
and p53 GRNs are shown in Tables
3.2
and
3.3
respectively (these are kept hidden
from our algorithm).
Our method uses an open-source and efficient exact SAT-solver, MiniSAT
v1.14 [
17
]. Shell scripts were created to invoke MiniSAT and to implement the
All-SAT functionality. All tests were implemented and run on a Core 2 Duo Mac
OSX machine with 4 GB RAM. Runtimes depend on the input predictor set and
