Facilitated Variation: How Evolution Learns from Past Environments To Generalize to New Environments

Abstract

One of the striking features of evolution is the appearance of novel structures in organisms. Recently, Kirschner and Gerhart have integrated discoveries in evolution, genetics, and developmental biology to form a theory of facilitated variation (FV). The key observation is that organisms are designed such that random genetic changes are channeled in phenotypic directions that are potentially useful. An open question is how FV spontaneously emerges during evolution. Here, we address this by means of computer simulations of two well-studied model systems, logic circuits and RNA secondary structure. We find that evolution of FV is enhanced in environments that change from time to time in a systematic way: the varying environments are made of the same set of subgoals but in different combinations. We find that organisms that evolve under such varying goals not only remember their history but also generalize to future environments, exhibiting high adaptability to novel goals. Rapid adaptation is seen to goals composed of the same subgoals in novel combinations, and to goals where one of the subgoals was never seen in the history of the organism. The mechanisms for such enhanced generation of novelty (generalization) are analyzed, as is the way that organisms store information in their genomes about their past environments. Elements of facilitated variation theory, such as weak regulatory linkage, modularity, and reduced pleiotropy of mutations, evolve spontaneously under these conditions. Thus, environments that change in a systematic, modular fashion seem to promote facilitated variation and allow evolution to generalize to novel conditions.

How Can Evolution Learn?

“…First, a memory of phenotypes that have been selected in the past (e.g., Figure 1) can facilitate faster adaptation whenever these phenotypes are selected again in the future [8,33]. Second, and more importantly, because learned models can generalise (e.g., Figure 1J), an evolved memory can, as illustrated by Parter et al [8], also facilitate faster adaptation to new targets. In short, evolvability is to evolution as generalisation is to learning.…”

“…The evolved GRN thus forms an associative memory of phenotypes that have been selected for in the past, spontaneously recreating these phenotypes as attractors of development with the GRN and also producing new phenotypes that are generalisations of them. (A-D) A GRN is evolved to produce first one phenotype and then another in an alternating manner [8,49]: A = Charles Darwin, B = Donald Hebb (who first described Hebbian learning). The resulting phenotype is not merely an average of the two phenotypic patterns that were selected in the past (as per a univariate model or free recombination of phenotype pixels).…”

“…(E-J) In a separate experiment, selection iterates over a set of target phenotypes (E-H). In addition to developing phenotypes that match patterns selected in the past (e.g., I), this GRN also generalises to produce new phenotypes that were not selected for in the past but belong to a structurally similar class, for example, by creating novel combinations of evolved modules (e.g., developmental attractors exist for a phenotype with all four 'loops' [32], J) -see also [8]. This demonstrates a capability for evolution to exhibit phenotypic novelty in exactly the same sense that learning neural networks can generalise from past experience [32].…”

“…Sometimes implies an external teacher that knows the desired output -but in practice, it usually means simply modifying parameters by gradient descent on an error function (rather than by trial and error) [14]. The combination of random variation and selection can affect the same changes in a model given the same gradient, and thus an analogue of selection for a particular target phenotype or phenotypes [32] (see also selection in varying environments or for a set of target phenotypes [8,49]). …”

Stability and Flexibility from a System Analysis of Gene Regulatory Networks Based on Ordinary Differential Equations

“…In engineered systems, modules may be stable (in both our meaning and in the topological meaning [45]) subunits performing different tasks. In biological systems, modularity brings an evolutionary flexibility to the system, and recent studies suggest it origins from time-varying evolutionary goals [46,47], with each module serving its own special task. Despite the general consensus about modularity in networks describing cells, the concept of modules has been used in various ways (but still motivated from a high process similarity within the modules [43,48,49]).…”