Search and Learning in the Immune System: Models of Immune Surveillance and Negative Selection

Abstract

To protect our bodies against malicious antigens, our immune system needs to solve several difficult information processing problems. We propose algorithmic models of how the immune system finds intruding antigen as quickly and robustly as possible (a search problem) and how it reliably distinguishes proteins of foreign origin from normal host proteins (a classification problem). The analysis of these models provides novel qualitative and quantitative insights into the workings of the immune system, and yields predictions that can be tested experimentally.