different branch IDs in gcf and scfl for the same species tree

[HiranyaSudasinghe]

Hello,

I am interested in comparing the gcf and scfl for some nodes of a species tree that is interest to me. For example as shown here: https://www.robertlanfear.com/blog/files/concordance_factors.html

What I want to explore is how scf and gcf relative to branch length and branch support as the example here:

However, instead of -scf I am using -scfl option which is recommended in the documentation.

The commands I use are:

compute gene concordance factors

iqtree2 -t ${SPECIES_TREE} --gcf ${TREE_FILE} --prefix concord.gcf

compute site concordance factor using likelihood

iqtree2 -te ${SPECIES_TREE} -s ${CONCAT_FASTA} --scfl 100 --prefix concord.scfl

where ${SPECIES_TREE} is the species tree; ${TREE_FILE} are the gene trees and ${CONCAT_FASTA} is the supermatrix alignment.

When I explore the outputs of the above two analyses, however, I note that the branch IDs are different in the two cases. For example see below:
concord.gcf.scf.cf.branch

concord.scfl.cf.branch

Since the branch IDs are different I cannot visualize the relationship between gcf and scfl relative to branch length and branch support

I am not exactly sure why I am getting two different branch IDs in gcf and scfl outputs because I am using the same species tree.

I would be very much grateful, if you could let me know how to fix this?

Thank you very much!
Hiranya

[roblanf]

Hi Hiranya,

This is interesting, thanks for noticing this. @bqminh will be the person who might know how to fix the problem as asked, (and now you've pointed it out I'll need to double check my code for this https://iqtree.github.io/doc/recipes/concordance-vector.

However, even without that there is a fix. If you follow the code here: https://iqtree.github.io/doc/recipes/concordance-vector, you'll see that one of the trees I produce has the sCF and gCF labels on the same tree. This way you can easily compare them using the tree file itself. Just search for the bit that says "Each internal branch in gcf.cf.tree will be annotated like this:", and you should be able to recreate the necessary steps to do the same. That way you don't need branch IDs at all.

Rob

[HiranyaSudasinghe]

Dear Rob,

Thank you for your reply!

So, I followed: https://iqtree.github.io/doc/recipes/concordance-vector

and calculated gcf as follows:
iqtree2 -te concord.scfl.cf.tree --gcf ${TREE_FILE} --prefix concord.gcf.new

where concord.scfl.cf.tree is the tree output from: iqtree2 -te ${SPECIES_TREE} -s ${CONCAT_FASTA} --scfl 100 --prefix concord.scfl

Then I compared the branch IDs again and now they are consistent!

concord.scfl.cf.branch

concord.gcf.new.cf.branch

So, now it should be possible to compare the scfl and gcf as I wanted.

However, if I follow the iqtree documentation as:

compute gene concordance factors

iqtree3 -t concat.treefile --gcf loci.treefile --prefix concord

compute site concordance factor using likelihood with v2.2.2

iqtree3 -te concat.treefile -s ALN_FILE --scfl 100 --prefix concord2

I think it need to be checked why the branch IDs are different and how to fix it.

Thank you very much!
Hiranya

[roblanf]

@HiranyaSudasinghe, I've also double checked things and I find the same. With my code the branch IDs are consistent.

I suspect the key difference is that in my code I compute the species tree first, and then fix it with -te in both gcf and scf analyses.

In your code, when you are just supplying the species tree as a starting tree with -t, it's possible that parts of the tree might change or perhaps even if they don't then in the attempt to optimise the tree something else changes which affects the branch IDs.

You could try re-doing your code using -te for both steps, and see what happens. Also, try comparing the entire phylogeny with your original approach. Are there any differences?

Rob

[HiranyaSudasinghe]

Dear Rob,

In my second reply, I ran both scfl (first) and gcf (tree output from scfl as the input) using -te option and then the branch IDs are consistent. So, I think you are right that when I used -t option for gcf something may have changed.

In terms of the number of taxa, nothing has been changed in the phylogenies using both approaches.

Best wishes
Hiranya

[drelo]

Dear @HiranyaSudasinghe and @roblanf I experienced the same issue with a big dataset since I was using this tutorial . Just imitating this portion of code

# infer a concatenation-based species tree with 1000 ultrafast bootstrap and
#an edge-linked partition model
iqtree3 -s ALN_FILE -p PARTITION_FILE --prefix concat -B 1000 -T AUTO

# infer the locus trees
iqtree3 -s ALN_FILE -S PARTITION_FILE --prefix loci -T AUTO

# compute gene concordance factors
iqtree3 -t concat.treefile --gcf loci.treefile --prefix concord

# compute site concordance factor using likelihood with v2.2.2
iqtree3 -te concat.treefile -s ALN_FILE --scfl 100 --prefix concord2

and as a template, I used the astral tree both times but the nodes numbering differ in both trees. This was using a dataset of ~3000 UCEs, with close to 400 UCEs the node numbers matched.

iqtree3 -t astral3111.tree --gcf ninety3111.trees --prefix geneastral90 -T 4
iqtree3 -te astral3111.tree -s eightyfiveAtNinety.fas --scfl 200 --prefix siteastral90 -T 4

I am a complete newbie to concordance factors (which is a great tool with my dataset/results) so thanks for the advice/fix here.

Best

Andrés

[roblanf]

Hi Andres,

You will need to set the species tree with -te each time. That tells IQ-TREE that it cannot change the input tree, so the node IDs will stay consistent.

Rob

[drelo]

Thanks Rob, yes I imitated the solution as @HiranyaSudasinghe using the -te and feeding the site.treefile into the next step with the gene trees and it worked.

[roblanf]

Thanks all. I'm going to close this one now. To re-iterate for anyone finding this discussion, this is the expected behaviour. In other words - if you set a starting tree with -t then IQ-TREE can update the tree topology and the branch ID's will most likely change. If you need branch ID's to stay the same in different runs, you must fix the tree with -te.

Of course, I do understand that this feels like a nuanced point, particularly if you're not used to it. Hopefully the existence of this thread will help disambiguate things for people!

Rob