Mapping the possible: Developmental Coarse-Graining for Evolutionary Prediction

Abstract Development matters for evolution because it determines which phenotypes are accessible over a given timescale. We argue that extending the predictive horizon of evolutionary biology therefore requires capturing this accessibility at a level of detail appropriate to the timescale of interest, an operation known as coarse-graining. Classical quantitative genetics summarizes accessibility as standing variation, but this representation is only valid while the distribution of this variation remains approximately constant. For longer-term predictions, we propose exploiting a widely observed feature of developmental architecture: some components are deeply conserved whereas others evolve rapidly, providing a principled basis for successive coarse-grainings. At intermediate timescales, prediction thus requires representing the phenotypes accessible through small perturbations around a population's current state, and we review emerging evidence that this is empirically tractable. Over still longer timescales, the relevant representation becomes the full range of phenotypes that can be generated by the conserved aspects of developmental architecture. Building the tools to estimate this developmental hierarchy from empirical data may be the central contribution evolutionary developmental biology can make to a predictive theory of phenotypic evolution spanning micro- and macroevolutionary timescales.