Biomedical Engineering Reference
In-Depth Information
understand the relationship between extracellular stimuli, intracellular signaling
dynamics, and multicellular tumor growth. Simulation results can be validated
with in vitro or in vivo experiments, or verified with other theoretical studies. For
using this approach, an algorithm for linking molecular and multicellular scales is
indispensable and sometimes needs experimentally supported creative thoughts.
On the basis of the modeling works presented here, we argue that this type of
ABM is highly suited to modeling complex emergent behaviors of cancerous
systems, which are generated as an outcome of direct and indirect interactions
between large numbers of individual cells.
The molecular-multicellular ABM not only enables the monitoring of multi-
cellular dynamics in response to molecular changes, but also facilitates the
tracking of the fate of molecular components per cell and cell cluster as the entire
tumor system evolves. It is now possible to ascertain the cause of a specific tumor
growth pattern at the multicellular level by exploring the time-course history of
intracellular signaling profiles within individual cells. For example, the EMT
process has been studied by using the ABMs presented in [
48
]; it is very difficult
for an averaged population-based continuum model to study this process. It is also
noteworthy that the molecular-multicellular ABM has been employed in other
biomedical fields as well other than cancer, e.g., in epithelial cell study [
54
,
55
]
and in acute inflammation study [
56
-
58
], highlighting the promise of this type of
ABM in translating mechanistic knowledge into an integrated experimental and
computational framework.
There are a number of technical challenges in transitioning these ABMs to
biomedical/clinical practice. These include the more common issues such as
obtaining access to relevant data to validate simulation results and defining stan-
dards for model definitions. The most severe issue, however, is the compute
intensity associated with these discrete-based hybrid models. In modeling cancer,
it is generally accepted that the higher a model's spatial and temporal resolution,
the higher its compute power demand [
4
]. ABMs are generally too detailed to
simulate over a long period of time, particularly in a large, 3D domain. Paralle-
lizing the code and then running the model on a cluster of supercomputers is a
possible but not always practical solution that still may not resolve all the diffi-
culties in handling the enormous amount of experimental and clinical data. We and
others have begun to turn to hybrid, multiscale and multi-resolution modeling
[
6
,
12
,
59
], where multi-resolution means that cells at distinct topographic regions
are treated differently in terms of the modeling approach applied. This approach
has the potential to achieve discretely high resolution wherever and whenever
necessary to improve the model's predictive power, while at the same time
reducing compute intensity as much as possible to support scalability of the
approach to clinically relevant levels. By drawing on the strengths of the multi-
resolution approach, and integrating it into the next generation ABM models with
a hierarchy of processes at varying time and space scales, we can produce com-
putationally efficient models to simulate tumor progression, predict treatment
impact, and ultimately, be applicable in clinical practice.
Search WWH ::
Custom Search