Information Technology Reference
In-Depth Information
of the T
H
cell agents becoming activated depending on whether the avidity with
the APC agents is above the recognition threshold. Once activated, the T
H
cell
agents lose their chemokine receptor and start move randomly resulting in them
drifting away from the paracortex region. At the end of the simulation iterations,
the activated T
H
cell agents are noted. Even though the overall behaviour of the
agents in the model may be what is expected, it goes some way to justify the
model as the individual pieces of biology detail that has been used to build
it, combines to produce behaviour (i.e. movement and interactions of immune
agents) similar to that seen in real lymph nodes.
Due to the number of parameters that can be changed in the simulator, many
different experiments can be run to investigate different issues and effects relating
to the behaviour of the model. It is noted that a large numbers of parameters
can often hinder the experimentation and results gained from simulations such as
ours. However, some initial parameter investigations suggest that the behaviour
of the simulator is insensitive to appropriate changes in many of the parameters
such as the cellular space sizes and chemical space parameters. These parameters
can therefore be kept constant for experimentation into the degenerate receptors.
This leaves the simulator with only a small manageable subset of the parameters
described above (such as the recognition threshold, antigen receptor and T
H
cell
receptors) that have a real effect on degenerate recognition in the model. By
investigating the effects of these parameters, useful design principles for an AIS
algorithm employing similar parameters should become apparent.
As an example, we present the results from an experiment investigating the
patterns of 10 unique T
H
cell agents with 8-bit receptors that become activated
when the simulator is run separately with 20 different 16-bit antigens. For each
antigen, the simulator is run 50 times and the percentage of simulations in which
each T
H
cell agent becomes activated is calculated. The results are shown in
Table 1, where a blank entry means that the T
H
cell agent did not become
activated. The parameters used for this experiment were:
w
= 50,
h
= 50,
pre itns
= 100,
itns
= 500,
chem prod
= 25%,
chem max
= 500,
ag max
=1,
apc num
= 10,
ag num
20,
th num
= 10,
recog
=4and
aff
= R-contiguous bits.
The degeneracy of the T
H
cells can clearly be seen in the results as each
T
H
cell is reacting to different antigen ligands (see definition in section 2). We
can also see that each of the 20 antigens invokes a unique set (pattern) of T
H
cells to become activated. These sets are of different sizes for different antigens,
ranging from 2 to 6 T
H
cells being activated. It is interesting to note that the
2T
H
cells that become activated by Antigen 9 are also activated by Antigen 8,
but the sets differ as Antigen 8 also activated 2 more T
H
cells. The percentage
values for the T
H
cell activations can be seen as a sensitivity that the T
H
cell
has for the antigen. In general, the results highlight the ability of 10 randomly
generated degenerate detectors to collectively distinguish between at least 20
different patterns based on the pattern of response of the detectors. This shows
that our model contains degerate detectors capable of reacting in different ways
to different patterns, and is therefore a tool we can use for further investigations
into the properties degeneracy.
