Branch lengths of combined morph+genetic analyses

[DeondeJager]

Hi,

How can one interpret the branch lengths of a combined morphological + genetic maximum likelihood analysis in IQ-TREE (v2.3.3)?

For example, with a genetic data-only dataset (unpartitioned, site-based GRT+I+G model), the branch length unit would be expected number of substitutions per site, while for a morphological data-only dataset (MKv model), the branch length unit would be expected number of changes per morphological character.

But what would the unit of branch length be for a combined analysis? Does it represent something specific, or is it a more generalised representation of evolutionary time?

Any help or thoughts welcome!

Cheers,
Deon

[DeondeJager]

I figured this out, so will post the answer here and then close, in case someone has a similar question in future. Feel free to re-open if needed.

My IQ-TREE command was: iqtree2 -p charset_MolUnpart_Roan_MorphPart.nex -o Outgroup --runs 2 -T 8 -B 10000, where charset_MolUnpart_Roan_MorphPart.nex specifies an unpartitioned molecular alignment and a morphological character alignment partitioned into ordered and unordered characters. Thus, I had three partitions in total.

The branch length scale of the final combined analysis ML tree was 0.04. Because I used -p to specify the character set nexus file in my command, it also specified to IQ-TREE to allow each partition to have its own evolution rate, which is the edge-proportional or edge-linked-proportional partition model (number 2 here), as stated in the *.iqtree output file (attached). This means that "each partition has its own specific rate, that rescales its branch lengths". The rates for each partition are provided in the *.iqtree output for the run as the "speed":

Edge-linked-proportional partition model with separate substitution models and separate rates across sites

  ID  Model           Speed  Parameters
   1  GTR+F+I+G4     0.9777  GTR{5.5485,82.1877,3.98,1.10246,99.3521}+F{0.311512,0.266028,0.164798,0.257662}+I{0.378363}+G4{0.393591}
   2  ORDERED+FQ+ASC 6.5218  ORDERED+ASC
   3  MK+FQ+ASC      4.1963  MK+ASC

Thus, to get to the expected number of substitutions per site or per character for a particular partition (i.e. ID: 1 [molecular], 2 [ordered morph characters] or 3 [unordered morph characters] in the table above), one has to simply multiply the partition-specific speed/rate with the overall branch length scale (0.04 in this case). I checked and doing this gives similar values to the branch length scales compared to runs where the molecular and morphological analyses were conducted separately.

charset_MolUnpart_Roan_MorphPart.nex.iqtree.txt