A Path Integral Method for Analytically Tractable Inference of Evolutionary Dynamics

Understanding the forces that shape genetic evolution is a subject of fundamental importance in biology and one with numerous practical applications. Modern experimental techniques give insight into these questions, but inferring evolutionary parameters from sequence data, such as how an organism’s genotype affects its fitness, remains challenging. Here we present a method to infer selection from genetic time-series data using a path integral approach based in statistical physics. Through extensive numerical tests we find that our method exceeds the current state of the art in the successful classification of mutations as beneficial or deleterious in a variety of scenarios, while also yielding orders of magnitude improvements in run time. Our approach can also be extended to jointly infer other evolutionary parameters such as the effective population size and mutation rates.