Moran process

How do new traits, from drug resistance in cancer cells to cooperative behaviors in animals, emerge and spread through a population? Understanding the interplay of chance and selection in finite populations is a central challenge in evolutionary biology. The ​​Moran process​​ offers an elegant and powerful framework to tackle this question. It is a simple stochastic model that strips evolution down to its bare essentials—birth, death, and selection—in a population of a constant size, providing profound insights into the fate of new mutations.

This article explores the theoretical foundations and practical applications of this foundational model. It addresses the knowledge gap between the complex reality of evolution and the need for a tractable mathematical description. By examining the Moran process, we can quantify the odds of evolutionary change and understand the timescales over which it occurs.

First, in the ​​Principles and Mechanisms​​ chapter, we will dissect the rules of the Moran process, starting with neutral evolution driven purely by chance and then introducing the impact of fitness differences and selection. We will see how this simple framework can be extended to model the complex social dynamics of evolutionary game theory and how microscopic random events give rise to predictable, macroscopic patterns. Following this, the chapter on ​​Applications and Interdisciplinary Connections​​ will demonstrate the model's remarkable versatility, showing how the same mathematical principles explain the progression of cancer, the maintenance of our tissues, the spread of engineered genes, and the co-evolutionary arms race between species.

To understand the dance of evolution in a finite population, we need a stage and a set of rules. Physicists love simple models, not because the world is simple, but because simple models expose the deep principles that govern even the most complex phenomena. The ​​Moran process​​ is exactly such a model—a beautiful, minimalist caricature of evolution that reveals profound truths about chance, selection, and the fate of genes.

Imagine a tiny, isolated world: a single crypt in the lining of your intestine holding a handful of stem cells, a small bacterial colony in a petri dish, or an endangered species clinging to a remote island. The defining feature of this world is that its population size, let's call it $N$, remains constant. There's no room for expansion; for every birth, there must be a death.

The Moran process codifies this elegant constraint with the simplest possible rule. At each tick of a discrete clock, two things happen:

1. One individual is chosen to reproduce.
2. One individual is chosen to die and is replaced by the offspring of the first.

That's it. This sequence of a single birth followed by a single death ensures the population size stays fixed at $N$. The beauty of this model lies in its fine-grained view of time. Unlike other models, such as the famous ​​Wright-Fisher model​​, where entire generations are replaced in synchronous, discrete steps, the Moran process unfolds one event at a time. Generations are overlapping; a newborn enters a population of individuals of all different ages. This "asynchronous updating" gives us a high-resolution lens through which to watch evolution happen.

Now, suppose our population contains two types of individuals, let's call them $A$ and $B$. If the individual chosen to reproduce is type $A$ and the one chosen to die is type $B$, the number of $A$'s in the population, which we'll call $i$, increases by exactly one. If a $B$ reproduces and an $A$ dies, $i$ decreases by one. If the reproducer and the deceased are of the same type, the count remains unchanged. The composition of the population thus takes a step-by-step walk, a one-dimensional journey on the integers from $0$ to $N$. What guides this walk? That is the central question of evolutionary dynamics.

Let's start with the simplest case: what if there is absolutely no difference in the intrinsic quality of types $A$ and $B$? They are perfectly equivalent in their ability to survive and reproduce. This is the world of ​​neutral evolution​​, where the only force at play is pure, unadulterated chance.

In this neutral world, a choice of who reproduces and who dies is a fair lottery. Every one of the $N$ individuals has a $1/N$ chance of being chosen for birth and a $1/N$ chance of being chosen for death. Let's think through the consequences.

Suppose we have $i$ individuals of type $A$ and $N-i$ of type $B$. What is the probability that the number of $A$'s increases to $i+1$? This happens if an $A$ is chosen to reproduce (a probability of $i/N$) and a $B$ is chosen to die (a probability of $(N-i)/N$). Since these choices are independent, the total probability is their product:

What about the probability of decreasing to $i-1$? This requires a $B$ to reproduce (probability $(N-i)/N$) and an $A$ to die (probability $i/N$). The total probability is:

Look at that! The probability of taking a step up is exactly the same as the probability of taking a step down. Our population's composition is on a perfectly unbiased random walk. There is no agenda, no arrow, no preference. It is evolution as a drunkard's walk.

This random walk cannot go on forever. Eventually, by chance, it will hit one of two absorbing boundaries: either $i=0$, where the $A$ type goes extinct, or $i=N$, where it achieves ​​fixation​​, becoming the sole type in the population.

This leads to a question of stunning simplicity and depth: If a single new mutant, type $A$, appears in a population of $N-1$ individuals of type $B$, what is its chance of ultimately taking over the entire population? The answer is one of the most elegant in all of biology. The fixation probability for a single neutral mutant is exactly:

Why? Think of it this way. In this completely fair game of chance, every single one of the $N$ individuals present at the start has an equal opportunity to be the ancestor of the entire future population. Our single mutant is just one of those $N$ individuals. Its chance is therefore one in $N$. The vast majority of new mutations, even if they are not harmful, are simply lost to the statistical noise of birth and death.

Now, let's load the dice. What if type $A$ is not neutral, but has a slight advantage? Perhaps it metabolizes food more efficiently or is more resistant to disease. We can quantify this advantage with a number, its ​​relative fitness​​, which we'll call $r$. If the resident type $B$ has fitness $1$, our mutant $A$ has fitness $r > 1$.

We modify the rules of our Moran game just slightly. The death lottery remains fair—any individual can be the unlucky one chosen to be replaced. But the birth lottery is now weighted by fitness. The probability of an individual being chosen to reproduce is proportional to its fitness. An individual with twice the fitness is twice as likely to be chosen to have an offspring in any given step.

This seemingly small change has dramatic consequences. The probabilities of moving up or down are no longer equal. The probability of the number of $A$'s increasing ($P_{i \to i+1}$) becomes greater than the probability of it decreasing ($P_{i \to i-1}$). The random walk now has a ​​drift​​—a gentle but persistent push in the direction of increasing $A$.

The fate of a mutant is still a game of chance, but the odds have shifted. The exact fixation probability for a type with relative fitness $r$, starting with $i$ copies, is given by a beautiful formula:

This formula is a treasure trove of insight. Notice what happens when we set $r=1$ (the neutral case). The formula becomes the indeterminate form $0/0$, but using a little calculus (L'Hôpital's rule), we find that the limit as $r \to 1$ is precisely $i/N$. The general theory of selection contains the theory of neutral drift as a special case, a hallmark of a powerful scientific framework.

For a single beneficial mutant ($i=1, r > 1$), its chance of fixation is now greater than $1/N$. But it is far from guaranteed! For example, in a large population, the fixation probability for a mutant with a small advantage $s$ (where $r=1+s$) is approximately $s/(1+s)$ (or simply $s$ for small $s$). A mutant with a $1\%$ advantage ($s=0.01$) only has about a $1\%$ chance of fixing. Selection provides a powerful tailwind, but it cannot eliminate the brutal reality of random luck, especially when the mutant is rare.

So far, we've imagined fitness as an intrinsic, fixed property of an individual. But in the real world, an organism's success often depends on its social environment. A cooperative strategy might thrive in a population of cooperators but be exploited into extinction in a population of cheats. This is the domain of ​​evolutionary game theory​​.

The Moran process provides a perfect framework for exploring these social dynamics. We can define a ​​payoff matrix​​ that describes the outcome of interactions. For example, when an $A$ meets a $B$, $A$ gets a payoff of $a$ and $B$ gets a payoff of $b$. We can then map these payoffs directly to fitness. The fitness of an individual of type $A$, $f_A(x)$, now becomes a function of the current frequency of $A$, $x=i/N$.

This makes the dynamics incredibly rich and fascinating. The "drift" on our random walk is no longer constant. It changes as the population composition changes. If cooperators do well when rare (because they find other cooperators to help) but poorly when common (because they are overrun by free-riders), the drift will push the population towards a stable mix of both types. The simple, elegant Moran process is now capable of modeling the complex ebb and flow of social strategies, from the production of public goods by bacteria to the evolution of altruism in animals.

We have seen that at its heart, the Moran process is stochastic—it is a game of chance. For a small population, the outcome is highly uncertain. But what happens as we scale up to a very large population, $N \to \infty$? Does any predictable pattern emerge from the countless random birth and death events?

The answer is a resounding yes, and it is one of the most profound ideas in all of science. If we zoom out and rescale time so that we are watching over "generations" (where a generation is about $N$ birth-death events), the frantic, random jiggles of the population frequency begin to blur. They average out, and a smooth, deterministic trend emerges. The expected change in frequency per unit of this rescaled time gives us a differential equation.

For frequency-dependent selection, this limiting equation is none other than the famous ​​replicator equation​​. The replicator equation, a cornerstone of deterministic evolutionary theory, states that the growth rate of a given type is proportional to the difference between its fitness and the average fitness of the population.

This is a beautiful unification. The predictable, deterministic laws of evolution that we often use to describe massive populations are nothing more than the macroscopic average of innumerable, microscopic, random events. The stochastic world of individual births and deaths gives rise to the deterministic world of smooth evolutionary flows. The coefficients in the more advanced Fokker-Planck equation, which describes the evolution of the entire probability distribution, directly correspond to this average "drift" and the magnitude of the random "diffusion" or noise around it.

We have constructed a powerful and versatile tool, starting from the simplest possible rules. But our model has one crucial, hidden assumption: that the population is "well-mixed," meaning every individual is equally likely to interact with every other. This is like assuming the population lives on a ​​complete graph​​ where everyone is everyone else's neighbor. What happens when we relax this? What if individuals live in a structured world, a lattice or a complex social network, where they only interact with their local neighbors? As we will see, this spatial structure can have dramatic and surprising effects, sometimes acting as an ​​amplifier of selection​​ and other times as a ​​suppressor of selection​​, fundamentally changing the odds of evolution.

Having acquainted ourselves with the machinery of the Moran process, we now arrive at the most exciting part of our journey. Like a physicist who, having learned the laws of mechanics, can suddenly see the universe in the fall of an apple and the orbit of the moon, we can now see the signature of this simple birth-death process in an astonishing array of natural phenomena. The true beauty of a physical law or a mathematical model is not in its complexity, but in its unity—its ability to describe a vast range of behaviors with a single, elegant idea. The Moran process is just such an idea. We will see how it provides a common language to discuss the fate of cancer cells, the renewal of our tissues, the spread of engineered genes, the evolution of cooperation, and the endless arms race between hosts and their pathogens.

The most fundamental question in evolution is this: if a new trait appears in a single individual, what is its destiny? Will it vanish, a forgotten whisper in the annals of history, or will it rise to conquer the entire population? This event, the complete takeover by the descendants of a single mutant, is called "fixation." The Moran process gives us a remarkably precise answer to this question.

Let us imagine a population of constant size $N$. One individual acquires a mutation that gives it a relative fitness advantage, which we call $r$. If $r=1$, the mutation is "neutral"—it's no better or worse than the original. In this case, the game is one of pure chance. Each of the $N$ individuals has an equal opportunity for its lineage to be the one that, by sheer luck, eventually dominates. The probability of fixation is thus simply its initial frequency: $1/N$. A depressingly small number for large populations!

But what if the mutation confers a slight advantage, say $r > 1$? The answer, derived directly from the rules of the Moran process, is a jolt of lightning. The probability of fixation, $\rho_1$, for a single mutant is not $1/N$, but $\rho_1 = \frac{1 - 1/r}{1 - 1/r^N}$ This single formula is a Rosetta Stone for evolutionary dynamics. For a large population $N$, the term $1/r^N$ in the denominator becomes vanishingly small. The fixation probability then simplifies to a stunningly simple approximation: $\rho_1 \approx 1 - 1/r$. If we write the fitness advantage as a selection coefficient $s = r-1$, so $r = 1+s$, this probability is approximately $s/(1+s)$. For a small advantage $s \ll 1$, the chance of fixation is approximately $s$. This is a revolutionary result! A mutant with just a $1\%$ advantage ($s=0.01$) has a fixation probability of about $1\%$, which can be thousands of times greater than the neutral probability of $1/N$ in a large population. Selection is not a gentle guide; it is a powerful amplifier.

We see this drama play out in the microscopic theater of our own bodies. Consider a population of cancer cells under treatment, or an endometriotic lesion under chronic stress. If a single cell acquires a mutation that grants it resistance to a drug or a better ability to survive, its fitness $r$ is greater than 1. This formula tells us its chance of taking over the lesion and rendering the treatment useless. The grim mathematics of the Moran process governs the evolution of drug resistance, a central challenge in modern medicine.

The same mathematics, however, also describes the elegant process of tissue renewal. Our tissues are maintained by stem cells, which reside in special environments called "niches." A niche can be thought of as a collection of a fixed number of "seats," $N$, for stem cells. When a stem cell divides, one of its daughters must replace another stem cell, which is precisely the Moran process. A mutation that gives a stem cell a higher propensity for self-renewal or niche occupancy gives it a fitness advantage $r > 1$. Its probability of eventually colonizing the entire niche is given by the very same formula. This is how a single mutated stem cell can become the seed for a cancerous tumor.

The power of this idea extends beyond natural evolution. In the field of synthetic biology, scientists are designing "gene drives" to alter entire populations of wild organisms, for instance, to make mosquitoes incapable of transmitting malaria. A gene drive is a genetic element that cheats in the lottery of inheritance, ensuring it is passed on to more than its fair share of offspring. This "super-Mendelian" inheritance can be modeled as a powerful fitness advantage $r$ in the Moran process framework. The model allows us to calculate the probability that releasing just a few engineered individuals can lead to the drive spreading through the entire target population, a calculation of immense ecological consequence.

The question of "if" a mutant will fix is only half the story. The other half is "when." Evolution proceeds at a certain tempo, and the Moran process allows us to measure its beat.

Let's return to our stem cell niche, but this time, assume all cells are neutral ($r=1$). Even with no selection, the population is not static. Due to the random nature of birth and death, lineages are constantly going extinct by chance, until eventually, only one ancestral lineage remains. This state is called "monoclonality." How long does this process of "clonal turnover" take? Using the Moran process, one can derive that the expected time to reach monoclonality, starting with $N$ distinct lineages, is remarkably simple: it is proportional to the square of the population size ($N^2$). This means larger stem cell pools maintain their clonal diversity for longer, providing robustness to the tissue. In contrast, tissues with small, rapidly dividing stem cell populations, like the lining of our gut, will experience rapid clonal turnover.

Now, let's reintroduce selection and add another crucial ingredient: mutation. Beneficial mutations don't just appear on command; they are rare, random events. The time it takes for a population to adapt to a new challenge—like a pathogen evolving to escape the host's immune system—is the time spent waiting for the right mutant to appear and then successfully fix. This is a two-part problem: the rate of supply of new mutations, and the probability of success for each one.

Imagine a pathogen population of size $N$, where each birth has a small probability $\mu$ of producing an escape mutant. The total number of mutants generated per "generation" (roughly $N$ birth-death events) is proportional to $N\mu$. Each of these mutants has a fixation probability $\rho_{\text{fix}}$, which we know from our formula. The rate of successful mutations is therefore the product of the supply rate and the success probability. The expected waiting time for the first successful escape is simply the reciprocal of this rate. This leads to another profound insight: for large populations, this waiting time is inversely proportional to $N$. Larger populations evolve faster, not just because they have more individuals, but because they are a richer source of the winning lottery tickets that drive adaptation.

So far, we have imagined our populations as "well-mixed," like balls in an urn. But what if we add more structure? The robustness of the Moran process is that it can be extended to model these complexities, revealing even deeper truths.

​​The Importance of Place.​​ In many biological tissues, like the crypts of our intestines, location is everything. An intestinal crypt can be modeled as a one-dimensional line of cells, with a protected stem cell niche at the base. Only cells within the niche of size $N_b$ follow the Moran process of self-renewal. Cells outside the niche are transient; they divide and are pushed steadily upwards, eventually to be shed. What is the fate of a mutant? If it arises outside the niche, its lineage is simply washed away—its fixation probability is zero. For the crypt to be taken over, the mutation must occur in a stem cell within the niche. The fate of a mutation is not just determined by its fitness, but by its address. This spatial structure acts as a powerful firewall, protecting the tissue from the takeover of rogue cells.

​​Evolution as a Game.​​ In many scenarios, an individual's fitness is not a fixed property but depends on the behavior of others. This is the domain of game theory. The Moran process can be beautifully coupled with game theory to explore the evolution of social behaviors like cooperation. Imagine a population playing a "public goods game," where cooperators pay a cost $c$ to create a common benefit, and defectors pay no cost but reap the rewards. Can cooperation survive? By mapping the game payoffs to fitness and embedding the dynamics in a Moran process, we can calculate the fixation probability of a single cooperator in a population of defectors. The result depends on the population size, the group size of interaction, the benefit of cooperation, and the intensity of selection. This provides a bridge from the microscopic rules of interaction to the macroscopic fate of social behavior, linking biology to economics and the social sciences.

​​The Red Queen's Race.​​ Perhaps the most magnificent application is in modeling co-evolution, the endless arms race between species, such as a host and a pathogen. Here, we can model each population with its own Moran process, but with a twist: the fitness of a host type depends on the frequency of pathogen types, and vice versa. The fitness landscape of one population is the other population. What happens? Deterministic models often predict the system will settle into a stable equilibrium. But the stochastic Moran model reveals a different, more dynamic truth. The inherent randomness of birth and death in finite populations—the "demographic noise"—can continuously push the populations away from equilibrium, sustaining endless cycles of adaptation and counter-adaptation. The host evolves a defense, the pathogen evolves a counter-defense, and so on, in a chase that never ends. This is the "Red Queen" hypothesis, and the Moran process shows how demographic stochasticity itself can be the engine that drives it. It's a beautiful revelation: in the messy, noisy reality of biology, random chance is not just a nuisance to be averaged away; it is a fundamental creative force that can shape the grand tapestry of evolution.

From the cancer ward to the depths of our tissues, from the design of novel ecosystems to the origins of our social instincts, the simple rule of "one is born, one dies" provides a unifying thread. The Moran process, in its elegant simplicity, gives us a powerful lens to understand the interplay of chance and necessity that governs the living world.