Modeling Influenza Viral Dynamics in Tissue - Artificial Immune Systems - page 33

depict the dynamics. These sub-models could either be agent-based simulations
or ordinary differential equations when the spatial distribution of the agents
involved is not critical.
We also would like to refine the process of viral absorption, which is currently
described by the parameter k . It has recently been shown [28,29] that suscep-
tibility to a particular influenza strain is different depending on the cell type.
For example, human influenza virions preferentially bind to sialic acid (SA)- α -
2,6-Gal terminated saccharides found on the surface of ciliated epithelial cells of
the upper respiratory tract while avian influenza H5N1 prefers (SA)- α -2,3-Gal
found on goblet cells in and around the alveoli [28,29]. One easy way to take this
type of heterogeneity into consideration would be to define a virion absorption
rate rather than an infection rate, and consider different production rates, g V ,
for each strain of virus and for each cell type. Eventually, the parameter for
the absorption rate of virions, for example, could be broken into a sub-model
describing the molecular processes involved in virion absorption which would
explain in which way virus strains and cell receptors affect its value.
Eventually, when mechanisms such as viral absorption and release have been
modified to take on the form of molecular sub-models, the ABM will be calibrated
against a few different known influenza strains. This will provide pointers as
to which characteristics of an influenza viral strain drive these mechanisms.
Ultimately, we hope to be able to take a newly isolated influenza strain, infect
our in vitro system, and then fit our ABM to the experimental results. Doing
so would reveal the value of the parameters characterizing this particular strain
and hence reveal the lethality and infectivity of that strain.

6 Simulation Platform

The model is implemented on the MASyV (for Multi-Agent System Visualiza-
tion) simulation platform. MASyV facilitates the visualization of simulations
without the user being required to implement a graphical user interface (GUI).
The software uses a client-server architecture with the server providing I/O and
supervisory services to the client ABM simulation. The MASyV package con-
sists of a GUI server, masyv , a non-graphical command-line server for batch runs,
logmasyv , and a message passing library, ma message , containing functions to
be used by the client to communicate with the server. The simulation framework
iswritteninCandwasdevelopedonaLinux(Debian)system.
With the MASyV framework, a user can write a simple two-dimensional client
program in C, create the desired accompanying images for the agents with a paint
program of her/his choice (e.g. GIMP), and connect the model to the GUI us-
ing the functions provided in the message passing library. The flexible GUI of
MASyV, masyv , supports data logging and visualization services, and it supports
the recording of simulations to a wide range of video formats, maximizing porta-
bility and the ability to share simulation results collaborators. The GUI, masyv ,
is built using GTK+ widgets and functions. For better graphics performance,
the display screen widget, which displays the client simulation, uses GtkGLExt's