Abstract
Describing the evolution of a population of genomes evolving in a complex fitness landscape is generally very hard. We here introduce an approximate Gaussian closure scheme to characterize analytically the statistics of a genomic population in the so-called Quasi--Linkage Equilibrium (QLE) regime, applicable to generic values of the rates of mutation or recombination and fitness functions. The Gaussian approximation is illustrated on a short-range fitness landscape with two far away and competing maxima. It unveils the existence of a phase transition from a broad to a polarized distribution of genomes as the strength of epistatic couplings is increased, characterized by slow coarsening dynamics of competing allele domains. Results of the closure scheme are corroborated by numerical simulations.
Introduction
Predicting the genetic trajectory of a population is a fundamental challenge in evolutionary biology, particularly when fitness is determined by complex interactions between genes, known as epistasis. Standard models often rely on the assumption of Linkage Equilibrium (LE), where alleles at different loci are inherited independently. However, real-world fitness landscapes are frequently "rugged," where specific combinations of mutations yield higher fitness.
The Quasi-Linkage Equilibrium (QLE) regime accounts for correlations between genes but has historically been mathematically tractable only under the assumption of extremely high recombination rates ($r$). This research introduces a Gaussian closure scheme (GCL) to bridge this gap, providing a robust analytical framework that remains valid across a broader range of mutation and recombination rates.
Methodology
The authors employ a Gaussian closure scheme to simplify the infinite hierarchy of equations describing genomic moments (means, variances, and higher-order correlations). By tracking the first two cumulants—the average allele frequency at a site ($\chi_i$) and the pairwise covariance between sites ($\chi_{ij}$)—and assuming all cumulants of order three or higher are zero, the population is effectively modeled as a "cloud" in sequence space.
This approach yields a set of $\frac{1}{2}L(L+1)$ deterministic differential equations describing the evolution of these moments in the infinite-population limit. The validity of this Gaussian Ansatz was tested against stochastic numerical simulations of finite populations ($N=10,000$) using synthetic fitness landscapes, including random epistatic landscapes and a circular chromosome model with short-range interactions.
Key Insights and Results
The Gaussian scheme remains accurate as long as the total correlation received by a single site from all others is less than unity ($L \times |\chi_{ij}| < 1$), significantly outperforming standard QLE approximations.
Phase Transitions and Domain Coarsening
A central finding is the identification of a phase transition between "disordered" (paramagnetic) and "polarized" (ferromagnetic) population states. At high mutation rates, the population shows no clear preference for specific genomic configurations. As epistatic coupling increases or mutation rates decrease, the population enters a polarized phase where a specific optimal genome eventually dominates.
In this polarized phase, the genome initially forms "domains" of alternating alleles. These domains behave similarly to physical systems, where domain walls slowly diffuse and merge—a process known as coarsening—until the entire population reaches a uniform state.
Redefining QLE
The study suggests that QLE is not merely a high-recombination phenomenon but a regime characterized by the exponential decay of correlations with genomic distance. This behavior can occur in both the disordered and polarized phases, provided the Gaussian assumption holds. Numerical simulations confirmed that the time required for a population to settle into a single polarized state scales with the square of the genome length ($L^2$), consistent with the Allen-Cahn theory of coarsening.
Limitations and Future Directions
While powerful, the Gaussian closure scheme has limitations. It breaks down in "multi-modal" populations split into distinct, distant clusters (clonal competition) or when epistatic interactions are so strong that correlations exceed the validity threshold of the closure.
Future work may involve extending this framework to a "Mixture of Gaussians" to handle multi-modal distributions. Additionally, the scheme offers potential for "inverse statistical mechanics," where underlying fitness landscapes of real biological populations could be inferred directly from observed genomic sequence data.
Summary updated