Theoretical Models of Immune Processes - Immunological Computation: Theory and Applications - page 33

the B or T cell a nities over the total set of B or T cells. An induced B cell
cooperates with an activated T cell if it engages its T cell receptor by either
acting as an antigen-presenting cell (APC) or anti-antibody (anti-idiotypic
interaction).

Dynamic behavior of T cell clones is similar to that of B cell clones; however,
T cell clones are only driven by antigenic peptides on APCs. Specifi c peptides from
antibodies expressed and produced by B lymphocytes are not considered due to
their low individual concentration and frequency. h ereby, a bounded dynamics
of T cell clones is attained if and only if their receptors are considered a part of the
idiotypic network.

2.3 Multiepitope Immune Network

h is approach tried to map the IN theory into a parallel distributed processing
(PDP) (Vertosick and Kelly, 1989). h ey argued that B lymphocytes (or lympho-
cyte clones) can act as the units that compose a PDP network, that is,

• Receive inputs (from APCs, antigens, and cytokines)
• Generate output (antibody)
• Remember antigenic specifi city
• Convert inputs (antigenic stimulation) into output (antibody secretion) in a
quantitative fashion

h e PDP IN architecture can be designed to be multilayered, where lymphocytes,
plasma cells, and the lymphocytes that produce anti-idiotypic antibodies are con-
sidered as input, output, and hidden units, respectively (Figure 2.4).
h e connection weights between two lymphocytes can be defi ned in terms of
its a nities toward one another. h e learning behavior of the immune system uses
an unsupervised, local (Hebbian) learning rule. h is model also includes cytokines,
which are responsible for the clonal expansion of the population and subsequent
alteration of the connection strengths of a PDP composed of clonal units. h e
simulated annealing technique is used to fi nd the lowest energy confi guration of
the PDP network, by altering the shape of the activation functions of the units.
Using this model, complex antigen patterns (consisting of multiple epitopes) can be
learned and stored within the network.

2.4 Modeling the Germinal Center
Germinal centers (GCs) are the sites (in the follicles of the secondary lymphoid
nodes) where antigen-stimulated B cells complete their a nity maturation process
(Berek et al., 1991; Casamayor-Palleja et al., 1997; MacLennan). Particularly, GC