Cortier, T. | Search bioRxiv

bioRxiv2021

On eco-evolutionary dynamics of phenologies in competitive communities and their robustness to climate change

Global changes currently cause temporal shifts in the favourable conditions for different phases of species life cycles. Phenologies characterizing temporal presence, may adapt through heritable evolution in response to these changes. Given a community context, this evolution may cause a change in the phenology overlap and thus a change of interspecific interactions such as competition. Using a model in which phenologies compete and coevolve, we study the conditions under which diversity emerges, as well as their annual distribution. We find that the environment richness (food quantity, light, pollinators...) and competition constrain the diversity and spread of phenologies. A robust pattern of phenologies distribution emerges consistent with Swedish flowering observations. Once a stable community is reached, we apply a progressive change in environmental conditions. We found that adaptation eventually restored diversity, but that the simulated change often led to numerous extinctions due to increased competition. The percentage of diversity lost depends on the speed of change and on the initial diversity. Phenologies already pre-adapted to the new environmental conditions drive the restoration of diversity after the change. We finally study a spatial version of the model in which local communities are organized along an environmental gradient. Pre-change, allowing dispersal decreases the local adaptation of phenologies to their local fixed environmental conditions. Dispersal however largely enhances the maintenance of biodiversity in changing environments, though its benefits are not homogeneous in space. Evolution remains the only rescue mechanism for southern phenotypes.

evolutionary biology↗

bioRxiv2017

Resolving neuronal population code and coordination withgradient boosted trees

Understanding how neurons cooperate to integrate sensory inputs and guide behavior is a fundamental problem of neuroscience. A large body of methods have been developed to study neuronal firing at the single cell and population levels, generally seeking interpretability as well as predictivity. However, these methods are usually confronted with the lack of ground-truth necessary to validate the approach. Here, using neuronal data from the head-direction system, we present evidence how gradient boosted trees, a non-linear and supervised machine learning tool, learns the relationship between behavioral parameters and neuronal responses with high accuracy by optimizing the information rate. Interestingly, and unlike other classes of Machine Learning methods, the intrinsic structure of the trees can be interpreted in relation to behavior (e.g. to recover the tuning curves) or to study how neurons cooperate with their peers in the network. As an example, we show how the method reveals a temporally shifted coordination in a thalamo-cortical circuit during wakefulness and sleep, indicating a brain-state independent feed-forward circuit. Machine learning tools thus open new avenues for benchmarking model-based characterization of spike trains.

neuroscience↗