The challenge of scale in molecular adaptation: Local searches in astronomical genotype networks

TL;DR

The paper addresses how evolution navigates astronomically large genotype spaces and why adaptation is largely local. It combines genotype-to-phenotype map theory with viral quasispecies insights to show phenotype abundance biases mutational flux away from rare phenotypes, challenging the classic fitness-landscape metaphor. Key contributions include mean-field GP-map reasoning (arrival of the frequent), analyses of neutral-network percolation, and empirical hierarchical genotype networks from Qβ phage that demonstrate local exploration around abundant centers. The findings imply that robustness and navigability emerge from phenotype size and network structure, enabling efficient adaptation without global peak discovery and having broad implications for understanding viral evolution and molecular design.

Abstract

The exploration of vast genotype spaces poses fundamental challenges for evolving populations. As the number of genotypes encoding viable phenotypes grows exponentially with genome length, populations can only explore a tiny fraction of these immense spaces, a fact consistently supported by empirical and theoretical evidence. Paradoxically, local, mutation-driven searches near abundant sequences allow populations to generate phenotypic improvements and functional innovations despite this immense search space. In this contribution, we integrate insights from viral evolution with theoretical expectations derived from genotype-phenotype maps to re-examine how high-dimensional sequence spaces shape evolutionary dynamics. In resolving the paradox, abundant phenotypes play a crucial role because their combinatorial weight biases evolutionary trajectories. We discuss how this bias, together with limited accessibility of fitness peaks, modifies traditional metaphors -- such as fitness landscapes -- and challenges standard notions of evolutionary optimality. Our results underscore that adaptation is predominantly local yet remarkably efficient, providing a unifying perspective on the coexistence of robustness, innovation, and constrained exploration in molecular evolution.

Figures (3)

• Figure 1: Neutral networks in the RNA sequence-to-secondary structure genotype-phenotype map. This grid provides a simplified representation of genotype networks, which in reality are far more complex and not planar. Each node represents a unique RNA sequence, and a single point mutation corresponds to a displacement to a neighboring node. All nodes of the same color (connected by thick black edges) represent RNA sequences that fold into the same secondary structure---i.e., different genotypes mapping to the same phenotype—and together constitute that phenotype's neutral network. Three phenotypes (yellow, brown, and blue) occupy most of genotype space. The size of a neutral network ($N_1 = 15$ for the yellow folding, $N_2 = 10$ for the brown folding, and $N_3 = 12$ for the blue folding) is the number of sequences that produce the same phenotype. Many other, much smaller neutral networks (shown in various shades of gray) correspond to much rarer foldings, produced by only a few sequences (at most one or two in this example). Mutations within a neutral network allow divergence without losing a successful phenotype (hence without risking a fitness loss). Neutral networks usually traverse genotype space (e.g. the blue folding spans from top to bottom) and they get in touch with many other large phenotypes at one point or another (dashed red edges).
• Figure 2: Historical illustration of evolving viral quasispecies. Our understanding of how viral quasispecies explore sequence space has evolved over time as empirical evidence has accumulated. ( a) In early fitness‐landscape models, evolutionary dynamics were primarily described in terms of fitness increases. A simulated adaptive walk (brown curve with sampled red and blue points) illustrates this view. A viral population is assumed to start at low fitness, with individual variants independently sampling the local landscape. Fitter variants are selectively amplified, allowing the population as a whole to climb toward higher fitness peaks. Stochastic effects may occasionally enable escape from local optima. (b) When most mutations are nonviable or incur large fitness losses, the simple fitness‐ascent picture becomes problematic. This motivated a shift in emphasis from fitness progression to neutral navigation. Neutral networks (colored curves) span genotype space, connecting widely different sequences that produce the same phenotype through successive neutral mutations. A quasispecies can drift along such a network (thick black curve, red dots) without loss of fitness, while continually sampling nearby phenotypes. Occasionally, a previously distant neutral network (dashed black curve) comes into close proximity, offering access to higher fitness. The population then shifts to and explores this new network (blue dots). (c) Recent empirical evidence suggests that viral quasispecies do not primarily rely on neutral network connectivity to navigate fitness landscapes. Instead, they generate a large fraction of all possible variants within a relatively broad mutational neighborhood (outer black circles) around the most abundant sequence, which is itself highly populated (blue dots). Neutral variants (red dots) are explored within the same radius, rather than preferentially. The schematic separation between neutral and non-neutral variation is shown for clarity; in reality, neutral networks are interwoven with non-neutral neighborhoods. Shifts in the dominant (reference) sequence occur frequently (twice in this illustration), after which the statistical structure of the explored neighborhood is rapidly re-established.
• Figure 3: An empirical genotype network. Each node in this network represents a unique sequence (haplotype) of a Q$\beta$ phage population; two haplotypes are connected if they differ in just one nucleotide. Data correspond to passage 60 of a population adapting to a temperature of $43^{\circ}$C. A full description of the deep sequencing data and the curation protocol can be found in seoane:2025. Node size is proportional to the logarithm of the abundance of the haplotype (the number of sequences identical to that haplotype), and color conveys the same information. The spring layout (which positions nodes by treating them as masses connected by springs and minimizing tensions) suggests community structure around a series of salient nodes. These seem to act as sources of viral replicas which, through mutation, populate a neighborhood around each network hub. The self-similar structure of the network is reflected, among others, in a robust power-law degree distribution seoane:2025.