Abstract
The Adaptive Immune System as a Learning Algorithm
Shishi Luo, Los Alamos National Laboratory
Our adaptive immune system has the remarkable ability to learn and remember how to neutralize most foreign pathogens that invade our body. One pathogen for which this system seems to fail is human immunodeficiency virus (HIV). Suppose we formulate the adaptive immune response to a pathogen as a search process on some space of molecular shapes. what properties might it have and what is it about HIV that makes this process less effective?
Satisfiability and Evolution
Andrew Wan, Tsinghua University
How does a beneficial trait, one that is determined by the interaction between many genes, become fixed in a population? We consider the interaction between n genes and their evolving genotype frequencies, with reproduction through recombination, as a process on a Boolean function: when the Boolean function is satisfied by an assignment (i.e., a genotype), a small evolutionary advantage is conferred. Our main result shows that for any Boolean function and populations of polynomial size (polynomial in the number of genes and the inverse probability of initial satisfaction), with very high probability every individual in the population will be satisfying after polynomially many generations.
The Dynamics of Complex Adaptation
When and how can populations acquire complex adaptations that need multiple mutations to provide a fitness benefit? At least for toy population models, we can find simple, universal approximations for how the dynamics of the process depend on both the genetics of the adaptation (the number of necessary mutations, their individual fitness costs and rates of occurrence, and the recombination rate among them) and the size and structure of the population. Qualitatively, the potential complexity of adaptations is generally highest in large populations undergoing an intermediate amount of recombination.