Information Technology Reference
In-Depth Information
CalculateOptimum(self)
// get all r-bit templates eg **1100**
templates = getTemplatesIn(self)
// see below
addHoles(templates)
// count number of proteins induced
FOR each leaf template eg ****0001
numInduced = 1
FOR each non leaf template
numInduced = SUM numInduced for children
// Note that undetectable includes self
undetectable = SUM numInduced for roots
holes = undetectable - self
optimum = (nonself - holes)/nonself
addHoles(templates)
// iterative process
while (size of templates growing)
for each template
// case 1 - children - if *11** and its
‚spouse' *01** are both part of self,
then logically so are the children
*11* and **10*
IF templates contain spouse THEN
add children
// case 2 - parents - if **11* and its
‚sibling' **10* are both part of self
then logically so are the parents
*11** and *01**
IF templates contain sibling THEN
add parents
Fig. 6.
Wierzchon's algorithm [12] for counting number of holes using rContiguous bits. The
theoretical optimum is the size of non self space minus the number of holes.
The template algorithm suggested by Wierzchon gives us a useful metric for
measuring diversity. Figure 8 shows the number of templates (per antibody or protein)
found in the different self sets, antigen sets and corresponding antibodies produced
randomly and by the gene libraries. As expected, low numbers of self or antigen
clusters give the lowest numbers of templates (i.e. highest degree of clustering). One
library gives diverse antibodies, close to, or even higher than, random creation in
nature. Two and three libraries give the reverse; much tighter clustering, especially
for the 2self-2antigens scenario. Recalling this is a point of high coverage (with less
than the maximum numbers of antibodies) this is suggestive of a reason for the
libraries' efficacy.
