Using Genetic Algorithms to

Explore Pattern Recognition

in the Immune System

DRAFT
December 18, 1992

COMMENTS WELCOME

Stephanie Forrest
Dept. of Computer Science
University of New Mexico
Albuquerque, NM 87131
[email protected]

Brenda Javornik
Dept. of Computer Science
University of New Mexico
Albuquerque, NM 87131
[email protected]

Robert E. Smith
Dept. of Engineering Mechanics
University of Alabama
Tuscaloosa, AL 35487
[email protected]

Alan S. Perelson
Theoretical Division
Los Alamos National Laboratory
Los Alamos, NM 87545
[email protected]

Abstract

We describe an immune system model based on a universe of binary strings. The model is directed at understanding the pattern recognition processes and learning that take place at both the individual and species levels in the immune system. The genetic algorithm (GA) is a central component of our model. In the paper we study the behavior of the GA on two pattern recognition problems that are relevant to natural immune systems. Finally, we compare our model with explicit fitness sharing techniques for genetic algorithms, and show that our model implements a form of implicit fitness sharing.

1 Introduction

Our immune system protects us from an extraordinarily large variety of bacteria, viruses, and other pathogenic organisms. It also constantly surveys the body for the presence of abnormal cells, such as tumor cells and virally infected cells, and destroys such cells when they are found. To perform these tasks the immune system must be capable of distinguishing self cells and molecules, which it should not destroy, from foreign cells and molecules (antigens), which it should destroy. The enormity of this task has not been fully quantified, but Inman [17] has calculated that the immune system appears to be able to recognize at least 1016 foreign molecules. In practical terms, essentially any foreign molecule presented to the immune system, even those created in the laboratory and thus never having appeared before in all