Search for: All records

The course of evolution is strongly shaped by interaction between mutations. Such epistasis can yield rugged sequence–function maps and constrain the availability of adaptive paths. While theoretical intuition is often built on global statistics of large, homogeneous model landscapes, mutagenesis measurements necessarily probe a limited neighborhood of a reference genotype. It is unclear to what extent local topography of a real epistatic landscape represents its global shape. Here, we demonstrate that epistatic landscapes can be heterogeneously rugged and this heterogeneity may render biomolecules more evolvable. By characterizing a multipeaked fitness landscape of a SARS-CoV-2 antibody mutant library, we show that heterogeneous ruggedness arises from sparse epistatic hotspots, whose mutation impacts the fitness effect of numerous sequence sites. Surprisingly, mutating an epistatic hotspot may enhance, rather than reduce, the accessibility of the fittest genotype, while increasing the overall ruggedness. Further, migratory constraints in real space alleviate mutational constraints in sequence space, which not only diversify direct paths taken but may also turn a road-blocking fitness peak into a stepping stone leading toward the global optimum. Our results suggest that a hierarchy of epistatic hotspots may organize the fitness landscape in such a way that path-orienting ruggedness confers global smoothness. 

To respond and adapt, cells use surface receptors to sense environmental cues. While biochemical signal processing inside the cell is studied in depth, less is known about how physical processes during cell–cell contact impact signal acquisition. New experiments found that fast-evolving immune B cells in germinal centers (GCs) apply force to acquire antigen clusters prior to internalization, suggesting adaptive benefits of physical information extraction. We present a theory of stochastic antigen transfer and show that maximizing information gain via physical extraction can explain the dramatic phenotypic transition from naive to GC B cells—attenuated receptor signaling, enhanced force usage, and decentralized contact architecture. Our model suggests that binding-lifetime measurement and physical extraction serve as complementary modes of antigen recognition, greatly extending the dynamic range of affinity discrimination when combined. This physical-information framework further predicts that the optimal size of receptor clusters decreases as affinity improves, rationalizing the use of a multifocal synaptic pattern seen in GC B cells. By linking extraction dynamics to selection fidelity via discriminatory performance, we propose that cells may physically enhance information acquisition to sustain adaptive evolution. 

We explore the connection between migration patterns and emergent behaviors of evolving populations in spatially heterogeneous environments. Despite extensive studies in ecologically and medically important systems, a unifying framework that clarifies this connection and makes concrete predictions remains much needed. Using a simple evolutionary model on a network of interconnected habitats with distinct fitness landscapes, we demonstrate a fundamental connection between migration feedback, emergent ecotypes, and an unusual form of discontinuous critical transition. We show how migration feedback generates spatially non-local niches in which emergent ecotypes can specialize. Rugged fitness landscapes lead to a complex, yet understandable, phase diagram in which different ecotypes coexist under different migration patterns. The discontinuous transitions are distinct from the standard first-order phase transitions in statistical physics. They arise due to simultaneous transcritical bifurcations and exhibit a “fine structure” due to symmetry breaking between intra- and inter-ecotype interactions. We suggest feasible experiments to test our predictions. 

Generative adversarial networks (GANs) are a class of machine-learning models that use adversarial training to generate new samples with the same (potentially very complex) statistics as the training samples. One major form of training failure, known as mode collapse, involves the generator failing to reproduce the full diversity of modes in the target probability distribution. Here, we present an effective model of GAN training, which captures the learning dynamics by replacing the generator neural network with a collection of particles in the output space; particles are coupled by a universal kernel valid for certain wide neural networks and high-dimensional inputs. The generality of our simplified model allows us to study the conditions under which mode collapse occurs. Indeed, experiments which vary the effective kernel of the generator reveal a mode collapse transition, the shape of which can be related to the type of discriminator through the frequency principle. Further, we find that gradient regularizers of intermediate strengths can optimally yield convergence through critical damping of the generator dynamics. Our effective GAN model thus provides an interpretable physical framework for understanding and improving adversarial training. 

Cells are known to exert forces to sense their physical surroundings for guidance of motion and fate decisions. Here, we propose that cells might do mechanical work to drive their own evolution, taking inspiration from the adaptive immune system. Growing evidence indicates that immune B cells—capable of rapid Darwinian evolution—use cytoskeletal forces to actively extract antigens from other cells’ surfaces. To elucidate the evolutionary significance of force usage, we develop a theory of tug-of-war antigen extraction that maps receptor binding characteristics to clonal reproductive fitness, revealing physical determinants of selection strength. This framework unifies mechanosensing and affinity-discrimination capabilities of evolving cells: Pulling against stiff antigen tethers enhances discrimination stringency at the expense of absolute extraction. As a consequence, active force usage can accelerate adaptation but may also cause extinction of cell populations, resulting in an optimal range of pulling strength that matches molecular rupture forces observed in cells. Our work suggests that nonequilibrium, physical extraction of environmental signals can make biological systems more evolvable at a moderate energy cost.