IEEE Technology and Society Magazine - June 2019 - 59

to an alteration of the cell-cycle - the process by which a
cell replicates its DNA and divides into two daughter
cells - producing an uncontrolled cell growth [1]. At the
molecular level, an individual cell can be described by
the network of its signal transduction pathways [15], [16].
Modeling this network allows the representation of the
molecular machinery by which a cell integrates environmental signals and alters the gene expression patterns in
response to specific stimuli. For example, when a growth
factor binds to a membrane receptor (the stimulus), this
signal is translated into an activation of the cell cycle, i.e.,
the cell starts to grow in order to divide into two daughter
cells. Models of signaling networks have been reconstructed for different cell types and can be simulated
using different approaches, such as Ordinary Differential
Equation or Boolean Formalism [16], [17]. Such models
describe the state of different signaling pathways, which
can be used to predict the "fate" of a particular cell (proliferation/apoptosis). In turn, these models can be used
to predict the effects of drug synergies, as in the case of

[15], where a model reconstructed and calibrated for the
gastric adenocarcinoma cell-line (AGS) was used to accurately predict (as validated through growth experiments)
the effects of drug synergies.
At the cell-population level, a simple Agent-Based
Model (see Figure 2(c)) has been developed that takes
into account gene expression levels (see Figure 2(a)),
growth, as well as nutrient consumption, and where
each individual agent has its own signaling network from
which the propensity, for an individual cell, to proliferate
or to enter into apoptosis can be calculated using Boolean simulations (see Figure 2(b)). This model is being
used to integrate and interpret experimental data on
gene variability and drug synergies. A fundamental feature of simulating cell populations is that it allows for
modeling heterogeneity. This is of importance in the
study of cancer, since tumor heterogeneity and resistance to target therapies are two closely interconnected
phenomena [18]. Integrating experimental information of
gene expression variability available from RNA sequence

Node State Inferred
From Gene Expression

Signaling Model Controls
Agent Internal State

(a)
Growing
Cell

N 0 Cells

Supply MEK Inhibitor
Supply PI3K Inhibitor

Resistent
Cell

20

60

Active
Active (P)
Active (A)

Inactive
Inhibits
Activate

(b)
Quiescent
Cell

Cell Agent Properties
- Size (Cell Volume)
- Position (X, Y)
- Internal Clock

t

- Cell Internal State:
∗ Cell Cycle Phase
∗ Expression Profile

t=0

t=5

t = 10

t = 20

t = 40

t = 50

t = 60

t = 80

t = 100

t = 150

Simulating Drug
Scheduling
in Heterogeneous
Populations

(d)

∗ Proliferation Propensity
∗ Apoptosis Propensity

(c)

FIGURE 2. The simulation process of the hybrid multi-scale model of individual cell signaling. (a) RNA-seq expression profiles for different
conditions (i.e., different drugs, controls) for each node of the model. (b) Signaling model calibrated for the gastric adenocarcinoma (AGS) cell
line. Nodes are set to active or inactive based on the RNA-sequence data. (c) Different signaling networks are embedded into an agent-based
model (ABM) to simulate population level dynamics. (d) In-silico experiments of cell cultures treated with different drugs.

JUNE 2019

∕

IEEE TECHNOLOGY AND SOCIETY MAGAZINE

59
IEEE Technology and Society Magazine - June 2019

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - June 2019
