How predictable is evolution?
If life was allowed to re-evolve on an identical earth, would we still see flowering plants, birds, and humans? Steven J Gould famously argued that chance (historical contingency) is so important in evolution that “replaying the tape of life” would result in an earth with life that is entirely alien to us. Indeed, it seems unlikely that the Age of Mammals could have happened without a chance asteroid wiping out the non-avian dinosaurs.
However, evolution is not driven by chance alone. Similar selective pressures have continued to produce similar forms. Dolphins, ichthyosaurs, and tuna all sport a fusiform body plan despite being distantly related –a phenomenon known as convergent evolution. Cases of convergent evolution act as “natural replay experiments” that allow us to study the predictability of evolution more closely.
Using cases of convergent evolution in cheilostome bryozoans, NOBLE will investigate how the origination of novelties, the adaptive landscape, and developmental biases impact evolutionary predictability.
Research questions
1a: Is the evolution of novelties predictable?
1b: Does increased specialization reduce the probability of trait loss?
2: Is phenotypic convergence explained by the adaptive landscape or chance?
3: Does trait integration constrain evolution?
Methodology
Like soldiers in ant colonies, avicularia in bryozoan colonies are individuals that are specialized for (presumably) defense. Avicularia have evolved independently in multiple bryozoan groups (a clear case of convergence) and show a wide variety of morphologies.
Using data from the literature, we will reconstruct ancestral states to determine how many times avicularia have evolved, and then calculate whether other traits (like the presence of alternate defensive structures) can predict those originations (Q.1a). This data will also allow us to determine how often avicularia have been lost, and whether more specialized avicularia are less likely to be lost (Q.1b)
To investigate the role of the adaptive landscape in convergent evolution, we will use machine learning and geometric morphometrics to generate avicularium morphospace. Then biomechanical modelling will be used to create a suite of plausible adaptive landscapes. Finally we will simulate evolution within the morphospace to determine whether chance (represented by Brownian motion) or the adaptive landscape (represented by biased random walks and Ornstein–Uhlenbeck) best explains the degree of phenotypic convergence between bryozoan groups (Q.2)
The last part of this project uses New Zealand fossil bryozoans at the NHM to explore developmental biases. We will quantify trends in in avicularium shape between ancestor-descent pairs 5 different levels of trait integration and phylogenetic relatedness. This will reveal whether trait integration influences evolutionary change, and whether this relationship is consistent between convergent traits (Q.3).