iqtree
very low gcf values on species trees #468
-
|
I am working on an empirical dataset of 2750 ultraconserved elements from mammals. I have ASTRAL species trees built with dense taxon sampling and well-vetted sequence data. However, calculations of gCF on this tree are routinely low. Among ingroup samples, most branch values are <60, a few are <10, and several are even less than 1. While the tree is known to contain many short internodes (likely due to rapid radiation), these values feel surprisingly low and widespread. I know that Minh et al. present some instances of very low gCFs despite many available loci but Im wondering if what I describe seems reasonable? |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 2 replies
-
|
I moved this to iqtree2 discussion, because it's not really an issue. The answer is: yes, this is very much expected. The gCF measures how many gene trees share a split seen on the focal tree (e.g. species tree estimated however). If you are using UCEs, gene tree estimation error will be huge, because a single UCE often contains very limited information. This is compounded by short branches in the species tree of course. In this case, concordance vectors are likely to be useful, because because the three ways of measuring concordance (genes, sites, and quartets) have different pros and cons, and can all give you useful and somewhat different information. Despite gene tree estimation error, all three measures can still tell you useful things! Matt Hahn and I wrote a review on this here: https://academic.oup.com/mbe/article/41/11/msae214/7824840 and there's an IQ-TREE tutorial on it here: https://iqtree.github.io/doc/recipes/concordance-vector Happy to keep the discussion going if you have more questions, but I'll close the issue for now. (And for future general discussion items, just stick with using the IQTREE2 discussion forum - we're about to fix it, we know it's a bit confusing right now, that's our fault!). Rob |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for the reply Rob! I did expect low concordance values due to the conserved nature of the loci and poorly resolved gene trees, but these seemed very low. One followup Q is that I thought gCF represented the proportion of possible gene trees concordant, meaning that even among my genes that have complete taxon sampling, only ~1% are supporting some splits. Maybe I just have lots of comblike gene trees :( I guess I could scan them for clade presence separately just as a sanity check. (And indeed we have been using the concordance tutorial which is a helpful resource!!) |
Beta Was this translation helpful? Give feedback.
-
|
Sorry for the slow follow up - 1% is definitely not very low for UCEs, particularly if it's a short branch. On short branches, most UCEs will have zero observed substitutions on that branch, meaning they have ~no chance of recovering it. Glad the tutorial is useful! |
Beta Was this translation helpful? Give feedback.
-
|
Gotcha. Thanks for the helpful perspective! |
Beta Was this translation helpful? Give feedback.
I moved this to iqtree2 discussion, because it's not really an issue.
The answer is: yes, this is very much expected.
The gCF measures how many gene trees share a split seen on the focal tree (e.g. species tree estimated however). If you are using UCEs, gene tree estimation error will be huge, because a single UCE often contains very limited information. This is compounded by short branches in the species tree of course.
In this case, concordance vectors are likely to be useful, because because the three ways of measuring concordance (genes, sites, and quartets) have different pros and cons, and can all give you useful and somewhat different information. Despite gene tree estimation error…