Chalmers Conferences, 9th European Conference on Mathematical and Theoretical Biology

Real time forecasting of near-future evolution: theory and experiments
Philip Gerrish

Last modified: 2014-04-01

Abstract


In contrast to expensive adaptive landscape approaches to predicting evolution, we explore an “adaptive process” approach. Instead of focusing on the fitness mountain to be climbed, adaptive process approaches focus on the mountaineer, i.e., the population, and how she climbs the mountain. The mountaineer’s gait and pace reveal something about the terrain on which she walks; similarly, a population’s composition carries information about the underlying adaptive landscape, about trends in that landscape, and even about what lies a few paces ahead. The adaptive process approach that we explore builds on classical theory, but where classical theory can predict the course of evolution over the course of a single generation, our approach in preliminary studies successfully predicts evolution tens of generations into the future based only on measurements taken from populations in real time.


Keywords


evolution; population genetics; statistical mechanics; cumulant expansion; distribution of fitness effects; prediction