Processes 2019, 7,37
4. Discussion and Emerging Applications
The immune system is made up of many interacting components that together drive a complex
spatio-temporal behavior during immune response. Thus, agent-based modeling is particularly
suitable for understanding the immune systems function in health and in disease conditions such as
cancer. Here, we reviewed the latest agent-based and hybrid models that investigate the contributions
of the immune system to cancer growth and the effect of immunotherapy. In this context, we focused
on models of immune-related tumor mechanobiology, tumor-associated vasculature, tumor-associated
lymphatics, tumor immunotherapies, tumor-enhancing immune cells, and finally, models focusing on
intra-tumor heterogeneity. Overall, ABM can generate novel hypotheses to be validated and refined
by future experiments. Development and refinement of multiscale agent-based models along with
experiments through an iterative process can improve our understanding of biological processes in
cancer and lead to the identification of novel prognostic and predictive biomarkers that can improve
therapies and help design and interpret the results of clinical trials [143].
Models investigating the tumor-enhancing effects of the immune system can provide useful
insights into managing tumor-immune interactions. Since the tumor microenvironment can be
very heterogeneous, care must be taken to appropriately model cell-cell interactions between cancer,
stromal, and immune cells, the extracellular matrix, and the secreted factors. Accurate data from
in vitro and in vivo experiments must be used to understand the transition from tumor-inhibiting to
tumor-enhancing immune cell types. In addition, since immune cells such as macrophages and T-cells
are usually recruited to the tumor by secreted factors, an evolving tumor vasculature is necessary
to accurately model these processes. Agent-based models of the tumor enhancing effects of the
immune system can help us better understand how to prevent or revert the immune system back
into a tumor-inhibiting phenotype. Thus, these models will help improve immunotherapies for
cancer treatment.
One limitation of the immunotherapy studies mentioned above is that although the models
are quantitative in many aspects, the modules governing drug delivery and response is relatively
qualitative or semi-quantitative in nature. This could potentially be resolved by combining
spatio-temporal agent-based models with traditional model types, such as Physiologically Based
Pharmacokinetic (PBPK) models to track drug distribution in different physiological compartments,
and pharmacodynamic (PD) models for individual cellular agents to represent effects of drugs on target
cells [144]. Such hybrid quantitative systems pharmacology (QSP) models can be utilized not only as
a platform for basic science research, but also as a potential complement to the clinical research and
drug discovery pipeline. A schematic of such a hybrid model based on the research in our own group
is presented in Figure 3. Compared with continuous models, the discretely represented agents allow
the flexibility to track intra-tumor heterogeneity, such as tumor neoantigen profile, T-cell clonality and
local expression of immune checkpoint molecules with preferred levels of granularity. By running
multiple simulations in parallel using different parameter values and initial conditions accounting
for genetic background and environmental exposure, the models can represent cohorts of patients
with desired population scale heterogeneity. These properties render agent-based and hybrid models a
powerful platform for conducting virtual clinical trials.
52
4. Discussion and Emerging Applications
The immune system is made up of many interacting components that together drive a complex
spatio-temporal behavior during immune response. Thus, agent-based modeling is particularly
suitable for understanding the immune systems function in health and in disease conditions such as
cancer. Here, we reviewed the latest agent-based and hybrid models that investigate the contributions
of the immune system to cancer growth and the effect of immunotherapy. In this context, we focused
on models of immune-related tumor mechanobiology, tumor-associated vasculature, tumor-associated
lymphatics, tumor immunotherapies, tumor-enhancing immune cells, and finally, models focusing on
intra-tumor heterogeneity. Overall, ABM can generate novel hypotheses to be validated and refined
by future experiments. Development and refinement of multiscale agent-based models along with
experiments through an iterative process can improve our understanding of biological processes in
cancer and lead to the identification of novel prognostic and predictive biomarkers that can improve
therapies and help design and interpret the results of clinical trials [143].
Models investigating the tumor-enhancing effects of the immune system can provide useful
insights into managing tumor-immune interactions. Since the tumor microenvironment can be
very heterogeneous, care must be taken to appropriately model cell-cell interactions between cancer,
stromal, and immune cells, the extracellular matrix, and the secreted factors. Accurate data from
in vitro and in vivo experiments must be used to understand the transition from tumor-inhibiting to
tumor-enhancing immune cell types. In addition, since immune cells such as macrophages and T-cells
are usually recruited to the tumor by secreted factors, an evolving tumor vasculature is necessary
to accurately model these processes. Agent-based models of the tumor enhancing effects of the
immune system can help us better understand how to prevent or revert the immune system back
into a tumor-inhibiting phenotype. Thus, these models will help improve immunotherapies for
cancer treatment.
One limitation of the immunotherapy studies mentioned above is that although the models
are quantitative in many aspects, the modules governing drug delivery and response is relatively
qualitative or semi-quantitative in nature. This could potentially be resolved by combining
spatio-temporal agent-based models with traditional model types, such as Physiologically Based
Pharmacokinetic (PBPK) models to track drug distribution in different physiological compartments,
and pharmacodynamic (PD) models for individual cellular agents to represent effects of drugs on target
cells [144]. Such hybrid quantitative systems pharmacology (QSP) models can be utilized not only as
a platform for basic science research, but also as a potential complement to the clinical research and
drug discovery pipeline. A schematic of such a hybrid model based on the research in our own group
is presented in Figure 3. Compared with continuous models, the discretely represented agents allow
the flexibility to track intra-tumor heterogeneity, such as tumor neoantigen profile, T-cell clonality and
local expression of immune checkpoint molecules with preferred levels of granularity. By running
multiple simulations in parallel using different parameter values and initial conditions accounting
for genetic background and environmental exposure, the models can represent cohorts of patients
with desired population scale heterogeneity. These properties render agent-based and hybrid models a
powerful platform for conducting virtual clinical trials.
52