species assumptions starBEAST

[Nathan Harness]

Hi, 

I am building a species-tree of about 20 species within a single genus using starBEAST. 

To do this, you identify which taxa (here, species) each individual is a member of. I am wondering what if this assumption is wrong? 

e.g. Say I assign 6 individuals as species 1 and six other individuals as species 2, but in reality they are all 12 the same species from the same population? 

Alternatively, What if I say 6 individuals are species 1, but in reality one or two are members of an entirely different species? 

Would there be indications of this on the output tree?

[Graham]

You could try the STACEY package. It uses the multispecies coalescent model (like StarBEAST) but also estimates species delimitations.

[Huw A. Ogilvie]

I think if you split up "good" species, the divergence times will be close to 0. If you lump "good" species, that model violation will be somewhat buffered by inflating the Ne parameter. Of course it's not ideal and the Ne estimates will no longer be reliable, but it's not as bad as you might think.

As Graham points out, the way to avoid this problem entirely is to use a multispecies coalescent method that jointly estimates the species tree and delimitation such as STACEY or BPP. This was done in a recent paper by Andermann et al. in Systematic Biology, "Allele Phasing Greatly Improves the Phylogenetic Utility of Ultraconserved Elements", which used STACEY instead of the more traditional delimitation followed by species tree inference two-step approach.