Inferring the evolutionary model of community-structuring traits with convolutional kitchen sinks: Code and data

Both methods perform well across a range of phylogenetically informative evolutionary models, but CKS outperforms MPD as tree size increases. We demonstrate CKS by inferring the evolutionary history of freeze tolerance in angiosperms. Our analysis is consistent with a late burst model of freeze tolerance, suggesting it evolved recently.

We suggest that data ordered on phylogenies such as trait values, species interactions, or community presence/absence are good candidates for CKS modeling because the generative models produce structured differences between neighboring points that CKS is well-suited for. We introduce the R package kitchen to perform CKS for generic application of the technique.

Methods

Data were simulated, processed, and analyzed in R. These analyses were performed to investigate the ability of two methods, a machine learning technique termed Convolutional Kitchen Sinks (CKS) and models trained on series of Mean Pairwise Distance (MPD) metrics, also termed MPD curve, to recover the evolutionary model of traits that communities are assembled on. Communities were simulated on both simulated and empirical phylogenies by evolving traits on the phylogenies according to an Early Burst transformation governed by a normally distributed evolutionary parameter.

Data were separated into training and test data. The evolutionary parameters used in simulation were then modeled as a function of the observed simulated communities in the training data, using both the CKS method and MPD method. The models were then tested by examining the relationship between the predicted and known parameters of the test data.

Finally, predictions using trained models were made on the known community of freeze-tolerant dicots.