🧬 biology

Mutual Information for a Replicator System: the Role of Mutation

This paper demonstrates that while mutation-free replicator systems lose memory of environmental sequences over time, the introduction of mutation paradoxically enables a population to retain and reconstruct the ordered history of past environments by facilitating the decoding of evolutionary flows.

Original authors: Yancheng Wang, Xinpei Shen, Natalia L. Komarova

Published 2026-10-23

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Yancheng Wang, Xinpei Shen, Natalia L. Komarova

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Life is often described as a struggle for survival, but it is also a continuous act of listening. Organisms do not merely react to the world around them; they absorb information about their surroundings and carry that knowledge forward. When a population of bacteria or a colony of insects faces changing conditions, their genetic makeup shifts to match the new reality. This process, driven by natural selection, suggests that a living group acts as a kind of archive, storing a record of the environments it has survived. But a fundamental question remains: how much of that history is actually preserved in the population's current state? Does the group remember the specific sequence of events that led it to where it is today, or does it only retain a vague summary of what it has experienced?

To understand this, scientists look at how populations change over time in a fluctuating world. They imagine a group of different types of organisms, each better suited to a specific environment. As the environment shifts—perhaps from a hot, dry season to a cool, wet one—the types that thrive change, and the overall mix of the population evolves. The core of this inquiry is whether the final mix of organisms can tell an observer exactly what happened in the past. Specifically, researchers want to know if the population remembers the very last environment it faced, or if it can reconstruct the entire ordered history of every environment it has encountered. A key variable in this story is mutation, the random copying errors that occur when organisms reproduce. Intuitively, mutation is seen as a source of noise, a force that blurs the past and erases clear signals. One might expect that more mutation would make it harder to read the history written in the population's DNA.

A team of researchers at the University of California, San Diego, investigated this question using a mathematical model of a replicating population. They tracked how the composition of a group changes as it moves through a series of distinct environments, such as different temperature regimes or nutrient sources. They measured the amount of information the final population composition held about the environments it had passed through. Their work revealed a surprising twist: the presence of mutation, rather than destroying memory, can actually enable the population to retain a precise, ordered record of its entire history.

When the researchers removed mutation from their model entirely, the results were stark. In a world without mutation, the population's composition acted like a simple sum of its past. If the environment changed back and forth between two states, the final mix of organisms depended only on how many times each state occurred, not on the order in which they happened. If the population spent ten epochs in a hot environment and ten in a cold one, the result was the same whether the heat came first or the cold did. As the number of environmental changes grew, the population's ability to identify the most recent environment vanished. The specific signal of the latest condition was drowned out by the accumulated weight of the past. The population retained a perfect count of how many times each environment appeared, but it lost the sequence. It was as if the group remembered the ingredients of a soup but had forgotten the order in which they were added.

However, the picture changed completely when the researchers introduced mutation. In this scenario, the population did not just sum up its experiences; it settled into a specific, stable state for each environment. If the environment remained constant for a long enough period, the population would evolve toward a unique equilibrium composition, a distinct fingerprint for that specific condition. The researchers found that if the environments were sufficiently different from one another, and if the time spent in each environment was long enough, the population could be identified with absolute certainty. By looking at the final mix of organisms, an observer could determine exactly which environment was present at that moment, regardless of what happened before.

Even more striking was the discovery that this clarity extended to the entire past. Because the population's movement toward a specific equilibrium is a reversible process in theory, the researchers showed that one could work backward from the final state. By identifying the last environment, they could mathematically reverse the evolutionary flow to find the state of the population just before that environment began. Repeating this step allowed them to reconstruct the entire ordered sequence of environments the population had experienced. This means that with mutation, the population does not just hold a summary of the past; it holds a detailed, chronological diary. The mutation, often thought of as a source of error, provided the mechanism that allowed the population to distinguish between different historical paths that would otherwise look identical.

The researchers confirmed these theoretical findings with computer simulations. They modeled a simple system with two types of organisms and two environments. Without mutation, the information about the latest environment dropped rapidly as the number of environmental changes increased, eventually becoming negligible. With a small amount of mutation, the information about the latest environment grew as the time spent in each environment increased, eventually reaching a level where the environment could be identified perfectly. When they tested the ability to reconstruct the full history, they found that while the mathematical logic allowed for perfect recovery of the entire sequence, practical limitations arose. As the history grew longer, the differences between the starting points of different histories became exponentially smaller. While the information was theoretically there, recovering it required extreme precision. Small errors in measuring the current population could amplify as one worked backward, making it difficult to distinguish the true history from a very similar one.

These results challenge the common view that mutation is purely a force of degradation. In the context of environmental memory, mutation acts as a stabilizer that allows a population to lock onto specific environmental states. Without it, the population's memory is limited to a count of events, losing the sequence entirely. With it, the population gains the ability to distinguish the order of events, preserving a complete record of its journey. The study highlights a subtle distinction between what is theoretically possible and what is practically accessible. While the population contains the full ordered history in its composition, reading that history becomes increasingly difficult as the record grows longer, especially if the observation of the population is not perfect.

This work provides a new way to think about how life interacts with its environment. It suggests that the genetic composition of a population is not just a snapshot of current fitness but a complex archive of its past. The ability to read this archive depends heavily on the biological mechanisms at play, particularly the role of mutation. Far from being a mere source of noise, mutation can be the key that unlocks the ability to remember not just what happened, but the precise order in which it happened. This insight deepens our understanding of evolution, showing that the mechanisms driving change also shape how the past is remembered and preserved within the living world.

Technical Summary: Mutual Information for a Replicator System: the Role of Mutation

Problem Statement

This research investigates the fundamental question of what information an evolving population retains about its environmental history. Specifically, it examines the mutual information between the sequence of environments experienced by a replicating population and the population's composition at a given time. The study focuses on a multispecies replicator-mutator system evolving in a fluctuating "fitness seascape" where environmental conditions change in discrete epochs. The central inquiry is to determine the extent to which the current population composition encodes the latest environment versus the entire ordered sequence of past environments, and how the presence or absence of mutation fundamentally alters this information retention.

Methodology

The authors employ a deterministic mathematical framework based on the quasi-species model (Eigen, 1971; Eigen and Schuster, 2012).

System Dynamics: The population consists of $S \ge 2$ species evolving in one of $M \ge 2$ environments. The dynamics are governed by ordinary differential equations (ODEs) describing the change in species frequencies $x_i$ . The fitness of species $j$ depends on the current environment $\varepsilon_r$ , denoted $f_{jr}$ .
Mutation: The system includes a fixed mutation matrix $Q$ $Q$ , where $Q_{ij}$ $Q_{ij}$ is the probability that a parent of species $j$ $j$ produces an offspring of species $i$ $i$ . The study compares two regimes:
1. No Mutation ( $Q = I$ ): The identity matrix, representing pure selection.
2. Strictly Positive Mutation: $Q_{ij} > 0$ for all $i, j$ , ensuring the system is irreducible.
Environmental Fluctuation: Environments change in epochs of fixed duration $T$ . The sequence of environments is modeled as an independent and identically distributed (iid) process, though the analysis of history recovery holds for general stochastic processes.
Information-Theoretic Metrics: The primary metric is Mutual Information (MI), $I(A; Z)$ $I (A; Z)$ , measured in bits. The authors analyze:
- $I(E_n; X_{n+1})$ : Information about the latest environment $E_n$ contained in the composition $X_{n+1}$ at the epoch's end.
- $I(H_L; X_L)$ : Information about the entire ordered history $H_L = (E_0, \dots, E_{L-1})$ contained in the final composition.
Analytical Tools:
- Log-ratio coordinates: Used to transform the simplex dynamics into a vector-valued random walk for the mutation-free case.
- Hilbert's Projective Distance: A metric on the positive simplex used to analyze contraction properties in the presence of mutation.
- Contraction Estimates: Leveraging the Perron-Frobenius theorem and properties of positive matrices to establish convergence to environment-specific equilibria.

Key Contributions and Results

1. The Mutation-Free Regime ( $Q = I$ )

In the absence of mutation, the system behaves as an additive accumulation of selection pressures.

Vanishing Matched Information: The mutual information between the latest environment and the population composition, $I(E_n; X_{n+1})$ , vanishes asymptotically as the number of epochs $n \to \infty$ . Specifically, it is bounded by $O(1/n)$ and is independent of the epoch duration $T$ under exact observation.
Exact Count Memory: Despite the loss of information about the order of environments, the population composition retains an exact record of the visit counts for each environment.
- Theorem 3.3: Under a necessary and sufficient condition of rational independence of the fitness difference vectors, the endpoint composition uniquely determines the number of times each environment occurred ( $C(L)$ ), though the temporal order is lost.
- Information-Theoretic Formulation: The conditional entropy $H(C(L) | X_L) = 0$ , meaning the history's count vector is perfectly recoverable from the composition. However, the order of events remains indistinguishable.

2. The Mutating Regime ( $Q > 0$ )

The introduction of strictly positive mutation fundamentally reorganizes how information is stored.

Saturated Matched Information: Unlike the mutation-free case, if epochs are sufficiently long ( $T > T_{suff}$ $T > T_{s u f f}$ ), the mutual information $I(E_n; X_{n+1})$ $I (E_{n}; X_{n + 1})$ saturates to the entropy of the environment, $H(E_n)$ $H (E_{n})$ .
- Mechanism: Positive mutation drives the system toward a unique, strictly positive equilibrium $x^*_r$ for each environment $\varepsilon_r$ .
- Separation: If the equilibria for different environments are distinct (guaranteed if fitness vectors are non-proportional and $Q$ is invertible), the population composition at the end of a sufficiently long epoch falls into disjoint "balls" in the Hilbert projective metric space centered at these equilibria.
- Robustness: This identification of the latest environment holds even under small bounded observation errors.
Exact Recovery of Ordered History:
- Theorem 4.6 & Corollary 4.10: Under the separation condition ( $T > T_{suff}$ ), the population composition at any finite horizon $L$ uniquely determines the entire ordered sequence of environments $(E_0, \dots, E_{L-1})$ .
- Reconstruction Algorithm: Because the flow map for a fixed environment is invertible, one can decode the latest environment, invert the flow to recover the previous state, and repeat recursively to reconstruct the full history.
- Contrast with Mutation-Free: This ordered memory is impossible without mutation, where only counts are preserved.

3. Numerical Illustrations and Practical Limitations

Numerical Sensitivity: While the mathematical reconstruction of the ordered history is exact for any finite horizon, the authors demonstrate that finite precision (floating-point arithmetic or measurement noise) poses a severe practical limit.
Error Amplification: The backward reconstruction algorithm involves inverting contraction maps. Small errors in the final composition are exponentially amplified as one steps backward through the history.
Results: Simulations show that while the latest few environments can be recovered with high accuracy, the success rate of recovering the full history drops rapidly as the backward depth increases, illustrating the distinction between exact mathematical recoverability and practical robust reconstruction.

Significance and Claims

The paper claims to provide a concrete information-theoretic characterization of evolutionary memory, distinguishing between:

Matched Information: Knowledge of the most recent environment.
Cumulative Memory: Knowledge of the aggregate history (counts).
Ordered Historical Memory: Knowledge of the specific sequence of events.

Key Insight: The authors argue that mutation, typically viewed as a source of noise that erodes genetic information, can paradoxically enable the retention of ordered environmental history. Without mutation, the system loses the ability to distinguish the order of events as the history grows (information about the latest environment vanishes). With mutation, the system's contraction toward distinct equilibria allows for the exact identification of the latest environment and, through invertibility, the reconstruction of the entire ordered past.

Modest Claims:

The paper does not claim that biological populations use this information for prediction or that retaining it improves fitness. The analysis is purely about the information present in the composition.
The results on ordered history recovery are strictly for exact observations at finite horizons. The authors explicitly note that under realistic conditions (finite precision, noise), the recoverable history length is limited by the amplification of errors during backward reconstruction.
The sufficient duration $T_{suff}$ derived for full recovery is a sufficient condition, not a necessary one; information may be recoverable for shorter durations if the specific support sets of the distributions happen to be disjoint, even if the theoretical "balls" overlap.

In summary, the work establishes that the presence of mutation transforms the population from a system that only remembers "how often" environments occurred to one that can, in principle, remember "in what order" they occurred, provided the observation is exact and the epochs are sufficiently long to allow convergence to distinct equilibria.