Title: Real-time forecasting of near future evolution: theory and experiment.

Abstract: I will present a cumulant expansion of a standard evolution model that I have been using to: 1) make inferences about the distribution of fitness effects (DFE) of newly-arising mutations, and 2) make short-term predictions about the future course of evolution. My colleagues from the University of Pennsylvania have done experiments with lab populations of E. coli to test this theoretical predictive framework, and so far the results on inferring the DFE are encouraging; the tests of forecasting accuracy are, well, more or less OK I guess, considering the dimensionality of the problem and other obstacles. I’ll show you data, DFE inferences, and predictions. I’m still trying to understand: 1) where this work sits in relation to classical population genetics, 2) where it sits in relation to other predictive frameworks in other fields, and 3) how you could actually use these results to predict evolution in the real world. For example, I’m still trying to figure out whether there is some sense in which I can claim that this framework is an extension of Fisher’s theorem and the Price equation (if so, then I can expect it to be very robust). So feedback will be greatly appreciated.