Multiple attractors in immunology: theory and experiment

Abstract

This selective survey discusses the relative merits of various modeling approaches in immunology that exhibit multiple attractors, and also assesses the ability of the different models to contribute to deeper biological understanding. The first topic is global anti-idiotypic network models, which, like Hopfield neural network models, exhibit a large number of steady states that are identified with memory. It is shown that a 'reverse engineering approach' to T-cell vaccination for autoimmunity, featuring steady states corresponding respectively to 'normality', 'vaccination' and 'disease', is able to spur new experiments, in spite of the model's deliberate neglect of almost all biological detail. Mention is made of several other T-cell models that feature bistability for Th1 or Th2 dominance, or for activation and unresponsiveness.