Artificial selection has generated over 300 horse breeds but attempts to domesticate zebras have repeatedly failed. What makes some organisms better at evolving than others? The answer must lie in differences in intrinsic properties that allow some organisms to generate “the right” type of heritable variation. Development can reveal these intrinsic properties, as indicated by evo-devo theory. However, we lack empirical methods to quantitatively describe this intrinsic capacity of organisms to generate novel phenotypic variation.
This project will change that. We will advance evo-devo theory into an empirical research program to predict the origin of new variation. For this, we will 1) formalize a modeling framework to predict variation by learning developmental rules directly from data, 2) apply this framework to underutilized empirical data of developmental trajectories, and 3) establish a complete pipeline, from data acquisition using modern technologies to modeling, that maximizes the predictive power of our framework. This pipeline will direct future empirical research and provide a unified approach to compare organisms in their ability to generate variation. By the end of the project, we will organize a workshop on data-driven prediction in biology and share results in four high-impact publications and conference presentations.
By establishing an empirical approach to model the generation of new phenotypic variation, we aim to overcome a critical bottleneck in our capacity to predict future evolution—a central objective in evolutionary biology. Our novel framework will thus have direct applications in various fields that require predicting or manipulating the direction of evolutionary processes, such as agriculture and medicine. In this way, the success of this project will provide concrete proof that there is much to gain from a serious account of organism-level processes like development in evolutionary biology and beyond.
