A branching diffusion model of selection: from the neutral Wright-Fisher case to the one including mutations
Abstract
We consider diffusion processes xt on the unit interval. Doob-transformation techniques consist of a selection of xt-paths procedure. The law of the transformed process is the one of a branching diffusion system of particles, each diffusing like a new process tildext, superposing an additional drift to the one of xt. Killing and/or branching of tildext-particles occur at some space-dependent rate lambda. For this transformed process, so in the class of branching diffusions, the question arises as to whether the particle system is sub-critical, critical or super-critical. In the first two cases, extinction occurs with probability one. We apply this circle of ideas to diffusion processes arising in population genetics. In this setup, the process xt is a Wright-Fisher (WF) diffusion, either neutral or with mutations. We study a particular Doob transform which is based on the exponential function in the usual fitness parameter sigma. We have in mind that this is an alternative way to introduce selection or fitness in both WF-like diffusions, leading to branching diffusion models ideas. For this Doob-transform model of fitness, the usual selection drift sigma x(1-x) should be superposed to the one of xt to form tildext which is the process that can branch, binarily. In the first neutral case, there is a trade-off between branching events giving birth to new particles and absorption at the boundaries, killing the particles. Under our assumptions, the branching diffusion process gets eventually globally extinct in finite time with exponential tails. In the second case with mutations, there is a trade-off between killing events removing some particles from the system and reflection at the boundaries where the particles survive. This branching diffusion process also gets eventually globally extinct but in very long finite time with power-law tails. Our approach relies on the spectral expansion of the transition probability kernels of both xt and tildext.
Turn this paper into a lesson
ArcXiv compiles a structured reading guide from this paper's metadata: plain-English importance, contributions, prerequisite concepts, which sections to read first, flashcards, and a quiz. Grounded in the abstract, never invented.