B Cell–Inspired Algorithms - Immunological Computation: Theory and Applications - page 122

h ese models consider interactions between antibodies and antigens as com-
plementary antigen-antibody matches or among antibodies themselves; however,
exact matching between antibody and antigen (i.e., between paratope and epitope)
is not used. Instead, some antibody-antigen matching measure (complementary
measure) is defi ned such that if its value is below some threshold, the antibody
does not react to the antigen at all. Binding between an antigen and an antibody
depends on how well an antibody's paratopes match an antigen's epitope; the closer
this match, the stronger the bind. Also, a similarity or a nity measure needs to be
defi ned to measure how well two antibodies match.
In case of AINs, antibodies and antigens need to be explicitly defi ned, which is
called a shape-space. Other elements are also to be considered in an AIN model,
which include
1. Number of antigen epitopes . Some models consider one epitope, others consider
several epitopes.
2. Number of antibody epitopes . Some models do not consider any epitope at all,
others consider one or several epitopes.
3. Number of antibody paratopes . Some models consider one or two paratopes.
4. Types of binding interaction between antibodies . Some models consider only
paratope-paratope interaction, only paratope-epitope or both paratope-
paratope interactions and paratope-epitope binding.
When an antibody has one paratope, it can bind to one epitope at a time; but, when
it has two paratopes, one paratope can bind to an antigen, whereas the other may
bind to an antibody with a similar epitope to the antigen.
Most AIN applications start with an input dataset that corresponds to a set
of antigens stimulating an immune network, which goes through a dynamic pro-
cess, until it reaches stability. Depending on the application, either the concen-
tration of each type of antibodies or the structure of the AIN or both are used as
results.
All models assume an initial confi guration of the IN; in some cases, the initial
confi guration of the IN is produced at random. h e IN undergoes a stimulation
process caused by the set of foreign antigens to the network; however, some models
consider analyzing the IN's intrinsic behavior when no antigens are present.
An IN is also represented as a graph, where nodes, edges, and arrows represent
antibodies, the interactions among them, (see Figure 5.4) and stimulation to the
AIN by foreign antigens, respectively.
IN models (De Boer, 1989) can be classifi ed into two categories: continuous-
and discrete models. In continuous models, the immune response is assumed to
be continuous, as opposed to discrete models where it is in discrete time steps.
Continuous models are described by a set of diff erential equations, and its purpose
is mainly in modeling biological phenomenon. However, discrete models are typi-
cally abstract functional models, which are called AINs, and its purpose is to solve
real-world computational problems.