CANCELLED - Evaluating the effects of poorly inferred trees on phylogenetic community structure

Background/Question/Methods Phylogenetic community structure (PCS) studies have become increasingly popular as a way to tease apart ecological and evolutionary processes assembling local communities. PCS studies combine a regional pool phylogeny and species lists of local communities to calculate metrics that summarize the relationships among co-occurring species. However, the vast majority of these studies take the phylogenetic tree as a given; only a handful of studies have investigated how poorly inferred trees impact PCS metrics. Here, we conduct a simulation study to understand the influence of phylogeny quality on PCS and inferences about ecological processes based on this type of analysis. To do this, we simulated 1000 trees, each containing 50 species (real trees). On these trees, we simulated DNA sequences of varying lengths (10-10,000 bp), and used these simulated sequences to produce new trees (estimated trees). We then calculated a measure of dissimilarity between each real-estimated tree pair. Two thousand communities were assembled from the pool of species in each real tree: 1000 under a competition scenario and 1000 under a neutral scenario. For each community, we estimated PCS using real and estimated trees, and correlated differences in perceived PCS with the dissimilarity between real and estimated trees.

Results/Conclusions Results show that conclusions about community structure are sensitive to quality of tree estimation. In our simulations, phylogenetic estimation needed at least 5,000 or 6,000 bp for PCS metrics calculated on estimated trees to approach those calculated on real trees. However, conclusions about the structure of the community were modified by poor trees only under the competition scenario. When communities were neutral, the quality of tree estimation seemed unimportant. This results suggest that the importance of neutral processes in studies of PCS can be overemphasized due to low quality phylogenies. On the other hand, for this type of analyses to detect a significant effect of competition (especially when this effect is weak), phylogenetic estimates need to be of very high quality.