---
title: Wright-Fisher Intra-Patch Reproduction
url: https://www.emergentmind.com/topics/wright-fisher-intra-patch-reproduction
type: topic
---

# Wright-Fisher Intra-Patch Reproduction

Wright-Fisher intra-patch reproduction describes the stochastic evolutionary dynamics of competing alleles or variants within a fixed-size population (“patch”), integrating the influences of genetic drift and natural selection. The formalism is anchored in the Wright-Fisher Fokker-Planck equation, enabling precise quantification of transition probabilities for allele frequencies over short timescales. Fisher’s angular transformation provides a natural way to regularize these dynamics, leading to closed-form Green’s functions that unify both neutral and selective scenarios and have direct applications in population genetics, evolutionary inference, and deep sequencing datasets [1508.01453].

## 1. The Wright-Fisher Fokker-Planck Equation in Fixed-Size Populations

The intra-patch regime assumes a fixed population size $N$, within which two or more self-reproducing variants compete. Let $G(x,x_{0};t)$ denote the transition density of the allele frequency $x \in [0,1]$ at time $t$ given initial frequency $x_0$. In the diffusion approximation, the Wright-Fisher Fokker-Planck equation is
\[
\frac{\partial}{\partial t}\,G(x,x_{0};t) = -\frac{\partial}{\partial x}\!\Bigl[s\,x(1-x)\,G\Bigr] + \frac{1}{2N}\,\frac{\partial^{2}}{\partial x^{2}}\!\Bigl[x(1-x)\,G\Bigr],
\]
with $G(x,x_{0};0)=\delta(x-x_{0})$. The “drift” term ($s$) encodes directional selection, while the “diffusion” term models genetic drift with variance $x(1-x)/N$. The equation is subject to absorbing boundaries at $x=0$ (extinction) and $x=1$ (fixation).

## 2. Fisher’s Angular Transformation and the Natural Length Scale

To simplify the coordinate-dependent diffusion, Fisher introduced the angular transformation:
\[
\theta = \cos^{-1}(1-2x),\qquad x = \tfrac{1}{2}(1-\cos\theta) = \sin^2(\tfrac{\theta}{2}).
\]
In $\theta$-space, the Jacobian $\frac{d\theta}{dx} = \frac{1}{\sqrt{x(1-x)}}$ regularizes the diffusion, mapping the frequency-dependent $D(x) = x(1-x)/N$ to a constant $1/N$ in $\theta$. This recasting allows direct analogy with Brownian motion, where the dynamics become coordinate-independent and more mathematically tractable.

## 3. Stochastic Differential Equation and the Effective Potential Landscape

In the angular coordinate, under neutrality ($s=0$), the transformation yields the SDE:
\[
\frac{d\theta}{dt} = -\frac{1}{2N}\cot\theta + \eta(t),\qquad \langle\eta(t)\eta(t')\rangle = \frac{1}{N} \delta(t-t').
\]
This drift can be expressed via the potential $U(\theta)=\frac{1}{2N}\ln(\sin\theta)$. The potential features a maximum at $\theta=\pi/2$ with divergences at boundaries ($\theta\rightarrow 0$ or $\pi$), corresponding to extinction or fixation. The allele frequency behaves as a Brownian particle in an unstable potential, highlighting the drift-driven tendency for rare variants to be lost and common variants to fix.

## 4. Short-Time Asymptotic Green’s Functions: Neutral and Selective Cases

### Neutral Case ($s=0$)
Expanding the drift around $\theta=\pi/2$ linearizes the process, leading to:
\[
\langle \theta(t) \rangle = \frac{\pi}{2} + (\theta_0 - \frac{\pi}{2})e^{t/(2N)},\qquad
\mathrm{Var}[\theta(t)] = e^{t/N} - 1.
\]
The Green’s function in $\theta$ and then $x$ space is
\[
G_x(x,x_0;t) =
\frac{1}{\sqrt{2\pi\,x(1-x)\,(e^{t/N}-1)}}
\exp\left[
-\frac{(\cos^{-1}(1-2x) - \cos^{-1}(1-2x_0)e^{t/(2N)} - \frac{\pi}{2}(1-e^{t/(2N)}))^2}{2(e^{t/N}-1)}
\right].
\]

### Arbitrary Selection ($s\neq 0$)
Here, the SDE in $\theta$ space is
\[
d\theta = -\frac{1}{2N}[\cot\theta - N s \sin\theta] dt + \eta(t).
\]
Using a heuristic Gaussian approximation:
- The mean $\langle \theta(t) \rangle$ solves the nonlinear deterministic ODE,
\[
\langle \theta(t) \rangle = \cos^{-1}\left[-\frac{1+2N\gamma \tanh(\gamma t/2 - \alpha)}{2N s}\right]
\]
with $\gamma = \frac{1}{2N} \sqrt{1 + 4N^2 s^2}$, $\alpha = \tanh^{-1}\big(\frac{2Ns\cos\theta_0 + 1}{2N\gamma}\big)$.
- The variance is determined by linearizing the drift around $\langle \theta(t) \rangle$,
\[
\mathrm{Var}[\theta(t)] \approx \frac{1}{2N\,\lambda(\langle \theta(t)\rangle)}(e^{2\lambda(\langle\theta\rangle)t}-1)
\]
where $\lambda(\theta) = \frac{1}{2N} [\csc^2\theta + N s \cos\theta]$.

The resulting $G_x(x,x_0;t)$ retains a closed Gaussian form.

## 5. Validity, Accuracy, and Regime Limitations

The heuristic Gaussian approximation assumes the process in $\theta$-space remains approximately Gaussian for short times ($t \ll$ mean fixation time), with the mean given by the full nonlinear ODE and variance by the local linearization. This regime applies for $t \ll \mathcal{O}(N)$ (neutral drift time scale) or $t \lesssim (1/s)\ln(Ns)$ (selected), and for frequencies not near absorbing boundaries. The approximation yields excellent agreement with stochastic simulation, accurately capturing fixation/extinction tails in these windows [1508.01453]. For longer times or frequencies near boundaries, accuracy necessarily degrades.

## 6. Implications and Applications in Evolutionary Genetics

Within a patch, allele frequencies evolve as Brownian particles in an unstable potential, providing an explicit physical interpretation for the eventual fate (fixation or extinction) of variants. The closed-form, short-time Green’s functions allow explicit computation of transition densities and time-dependent probabilities of fixation/extinction, unifying neutral and selective models in a single analytic framework. These results are particularly applicable to inference of selection from allele-frequency time series in high-throughput sequencing, especially in experimental microbial or viral populations with short generation times and large $N$, directly informing models of evolutionary dynamics under strong genetic drift and selection [1508.01453].

## 7. Connections and Extensions

The Fisher angular transformation not only regularizes the mathematical treatment of the Wright-Fisher process but also clarifies connections to other potential well models in stochastic processes. The asymptotic Green’s function framework developed by Hallatschek and colleagues establishes a rigorous foundation for time-dependent population-genetic inference. A plausible implication is that similar coordinate transformations could prove fruitful in other classes of stochastic, finite-population processes, particularly where strong boundary effects and short-time dynamics are central. 

For researchers analyzing deep sequencing or experimental evolution datasets, these intra-patch Wright-Fisher results provide the foundation for robust estimation of selection coefficients and evolutionary forecasting in finite populations [1508.01453].

Source: https://www.emergentmind.com/topics/wright-fisher-intra-patch-reproduction