Immunity-Based Computational Models - Immunological Computation: Theory and Applications - page 53

3.3.5 Landscape-Affi nity Matching

h is matching rule was proposed to capture the ideas of matching biochemical and
physical structures and approximate matching in the immune system (Harmer et al.,
2002). An input string and antibody strings are sampled as bytes and converted
into positive integer values to generate a landscape. Two strings are then compared
using a sliding window. In fact, three diff erent similarity measures are defi ned as
follows:

1. Diff erence-matching rule

N

∑

f X Y

i i

difference

i 1

2. Slope-matching rule

N

∑

f ( X X ) ( Y Y )

slope i 1 i i 1 i

i 1

3. Physical matching

N

∑

∀

f ( X Y ) 3 min( i , ( X Y ))

physical i 1 i i i

i 1

3.3.6 R-Contiguous Bits Matching

h e rcb matching rule, introduced by Percus et al. (1993), is defi ned as follows: If x
and y are equal-length strings defi ned over a fi nite alphabet, match ( x, y ) is true if x
and y agree in at least r contiguous locations.
As an example, if x

= ABA DCB AB and y = CAG DCB BA, then match ( x, y )

≤ >

is true for r 3. In the case of binary strings, a matching

rule typically used is rcb (Forrest et al., 1994) where a detector d is specifi ed by a
binary string c and threshold value r , and d matches a string x if rcb of c matches
the corresponding bits (at the same positions) of x . It was originally designed to
consider approximate matching between two strings. h e choice of rcb simplifi es
mathematical analysis and is a good model for approximate T cell matching. h e
parameter r determines a detector's degree of specifi city; the smaller the value of r ,
the more general is the detector.

3 and false for r