Further, we test this preliminary conclusion at the molecular-
cellular level. First, according to the modeling result, we summarize
the relative change of each agent from the normal liver to HCC in
Table2. Then, we summarize the relative change of each agent
from the normal liver to HCC in Table2 according to the experi-
mental data. The experimental results are collected from two inde-
pendent ways. One is the specialist knowledge which suggests the
current biological understanding of each agent and its relative
change from normal liver to HCC, relative changes of the agents
from clinical data has also been summarized as double check
[46, 47].The other way is to analyze the high-throughput micro-
array data, checking the transcriptional change of each agent from
the normal liver to HCC [52]. The comparison shows that the
modeling results and specialist knowledge have an agreement of
88%, and have an agreement of 62%, 76%, 71% with three indepen-
dent HCC tissues respectively (NCBI ID: GSE33006) [52].
The working endogenous network is established from the
interactions determined at different contexts and by different
groups. It seems quite striking that the model and experimental
results have a perfect match at the modular level, and are coherence
at the molecular-cellular level. Consistency of validation suggests
that the present working network of liver may reproduce the key
features of normal liver and HCC. Given the heterogeneous nature
of cancer and the incomplete working network, experimental
results of certain molecular-cellular agents that do not appear to
fit this model must be expected.
We also want to point out that as it is impossible to enumerate
all the possible attractors of this high-dimensional dynamic system,
the attractors obtained here are by sampling. By sampling enough
times, we conclude that there are at least five attractors; however,
we cannot preclude the possibility that there are other attractors.
Mathematically, the structure of the endogenous molecular-cellular
network model is similar to the Morse-Smale dynamical system
[53]: it is structurally stable with a finite number of attractors and
the transitions between attractors are possible in the presence of
perturbations.2.5 Model Prediction
Reproduces Key
Features of Cancer
Genetic Mutation Data
It has been known that dynamical equations can be used to predict
the effect of mutations. For example, a detailed analysis of muta-
tions against experiments has been performed for the core regu-
latory network of phage lambda genetic switch [54, 55]. Here, we
examined the molecular-level details of these two stable states not at
such dynamical level, but from the relative expression level side
which is less sensitive to kinetic parameters [56]. We first character-
ized the activity of each protein in a given stable state as activated or
inactivated by setting a threshold: if the activity of a protein is
greater than this threshold, we identified the protein as active,
and if lower, the protein was identified as inactive. ActivatedEndogenous Molecular-Cellular Network Cancer Theory 225