Systems Immunology: From Cells and Molecules to a Dynamic Multi-Scale System - Systems Biology Concepts and Insights - page 471

FIGURE 25.3 An integrative multi-scale computational model of the thymus using Statecharts. (A) A theory of interactions between
thymic epithelial cells and thymocytes presented as a Statechart. The visual multilayered nature of the language enables specification of rich
models emulating real biology. (B) A snapshot of the simulated thymus at run time. The visualization overlay on top of the executable program
enables intuitive understanding of the dynamics of the model and an interface for the modeler to alter parameters in the simulation. (Taken
from [65] )

commitment, the model showed that the thymus is packed
full of cells and competition exists between cells for inter-
action space. Suffice it to say that the rate of disassociation of
CD8 þ cells from epithelial cells is lower than that of CD4, to
create the observed unequal ratio. This is in contrast to the
two dominant molecular-basedmodels described above (see
Systems-focused section for more information on the
stochastic and instructive lineage commitment models). The
existence of competition has not been previously discussed,
and gaining the insight from in vivo experiments that cell
niche competition could be responsible for the observed cell
lineage ratio would have been difficult to deduce. Thus, the
framework provides an integrative multi-scale view that
enables us to capture emergent phenomena and matches the
way we reason in biology.

and accurate in-detail model of the immune system is likely
impractical in the foreseeable future. A big bottleneck lies
in the effort for accurate specification, which is currently an
extensive manual process requiring expert curators. Thus
the current models are either limited in the richness of the
biology they express (e.g., cellular automata) or rich in
their expression but narrow in scope (e.g., Statecharts). In
an era where gigabytes of new immune-related data are
being generated every day, this is a serious concern as the
former does not capture the reality being measured,
whereas the latter does not capture the scale. Solutions to
these concerns must advance on two fronts: first, as best as
possible, specification from the literature must become an
automated process. This will require advanced natural
language processing, automated ontology building, and
comprehensive high-quality standardized annotation of
genome-scale data, all fields still in their infancy. Begin-
ning with the cell

Limitations of a Multi-Scale-Focused
Approach
Owing to the high complexity of the system, its component-
rich multi-scale nature and the many unknowns, a full-scale

cytokine network, we have been
advancing this effort through immuneXpresso, an auto-
mated engine aimed at constructing an immune-related
knowledge base [67] . A second front is the incorporation of

e