Inferring the Evolutionary Model of Community-Structuring Traits with Convolutional Kitchen Sinks

Area of Science:

• Ecology
• Evolutionary Biology
• Computational Biology

Background:

• Community assembly processes, like trait filtering, can leave evolutionary signatures in community membership patterns.
• Understanding these signatures requires inferring the underlying trait evolution models.

Purpose of the Study:

• To introduce and evaluate Convolutional Kitchen Sinks (CKS) as a novel machine learning tool for inferring trait evolution models from community data.
• To compare CKS performance against traditional eco-phylogenetic metrics like mean pairwise phylogenetic distance (MPD).

Main Methods:

• Utilized community membership data to infer trait evolution models.
• Applied both MPD and the new CKS method.
• Tested methods across various phylogenetically informative evolutionary models.

Main Results:

• Both MPD and CKS successfully inferred evolutionary models.
• CKS demonstrated superior performance compared to MPD, especially with increasing phylogenetic tree size.
• Applied CKS to infer the evolutionary history of freeze tolerance in angiosperms.

Conclusions:

• The evolutionary history of freeze tolerance in angiosperms is consistent with a recent 'late burst' model.
• CKS is a powerful tool for analyzing diverse, phylogenetically ordered data, including trait values and species interactions.
• The R package 'kitchen' is introduced for general application of CKS.