[Feature Request] Allow ancestral state reconstruction for a strict consensus tree

[HS6986]

Dear Professor Smith,

I'm trying to reconstruct ancestral states using TreeSearch::PlotCharacter() for a strict consensus of trees generated by TreeSearch. However, TreeSearch::PlotCharacter doesn't seem to be able to reconstruct ancestral states well for a multifurcating tree; it appears to terminate without problems, but some of the reconstructed ancestral states look odd. I understand this is why Smith et al. (2024) used an arbitrarily selected most parsimonious tree for the ancestral state reconstruction.

So I have a suggestion. For each internal node in a strict consensus tree, we could examine how all most parsimonious trees reconstruct its state and then map them onto the strict consensus tree. For example, for a given node, if all most parsimonious trees reconstruct its state as -, we could map - to the node. Conversely, for another node, if some most parsimonious trees reconstruct its state as - while others reconstruct it as 1, we could map ? to the node.

I believe this feature would facilitate the interpretation of ancestral states. I would greatly appreciate it if you could consider implementing it in TreeSearch.

[ms609]

Thanks for the suggestion – this is an interesting idea. The problem I would anticipate is that the collapsed nodes may not be the only ones whose interpretation differs between trees: trees that reconstruct node A as $1$ may in turn reconstruct the parent/child of node A as $1$ rather than $-$. That said, it should not be too difficult to summarise this ambiguity across a set of trees.

[ms609]

I've had a first pass at implementing this. I've not had time to test it extensively, but perhaps you could see whether it works as you expect on your own datasets?

• Install the relevant branch with
  devtools::install_github("ms609/TreeSearch", branch = "179-consensus-reconstruction")

• Call PlotCharacter() with a list or multiPhylo object containing all most parsimonious trees.

This should also work in the EasyTrees() app - Select "Display → Characters on Trees → Tree to plot: 0"

[HS6986]

Thank you for attempting to implement my suggested feature.

I tried it with this version and it worked fine with both an R script and the EasyTrees() app for the dataset I am working on.

However, the EasyTrees() app always seems to show Error: subscript out of bounds when I do ancestral state reconstruction. This happens regardless of whether the tree to plot is a strict consensus or an MPT.

P.S. I've realized that it is probably because my dataset does not have character notes.

[ms609]

Thanks – a more informative error message to this effect would certainly be useful; I've added this to my list of improvements to make (#181)

[HS6986]

Thanks!