Time tree reconstruction - branch distances seem off

[alanxelena]

Hi everyone!

I've been using IQTree for a while now but this is my first attempt at using it to make a time tree reconstruction. I fed it with a core SNP alignment and a text file (tsv) with sample name and collection date as columns (Ranging between 2012 and 2020).
I got a result but I'm having a hard time understanding the branch lengths when I plot the tree using IToL: I attach a figure with a fragment of the generated tree as a sample.

My concerns are basically two: 1) The branch length (as date) seems to be very large - I asked IToL to display branch data as dates.
2) The upper branch, being longer, has a distance of 7 while the lower branch has a distance of 443.8. Shouldn't it be that longer branches represent a longer period of time?

Many thanks!

[tothuhien]

Hi @alanxelena , the branch lengths are not the dates. So the numbers you displayed in the plot are the actual dates of the corresponding nodes, not the lengths of the branches. I don't use iToL but could you select normal branch lengths as values to display instead of dates?

[alanxelena]

Hi! Thanks a lot for the very fast response. No, I understand the branch lengths to be given in a time unit (years maybe?). I can opt to use the distance instead (See attached). Plus, I can have the dates calculated (Attached, numbers below the line). For the leaves it works fine since that's the year I used in the date file but then it seems that the leaf year (for example 2013 in the upper clade) is substracted from the distance (71701,2) to infer the date at which the clade is formed (-69688,2).

[alanxelena]

I somehow thought it would be a good idea to reinstall iqtree3 from scratch, you know, "turn it on and off and it will work again".
Now I've run iqtree3 -s alignment.aln --date dates_test.txt and it crashed while looking for a model. This is not a big problem since I can manually input the best fitting one iqtree3 -s alignment.aln --date dates_test.txt -m GTR+F+ASC
but in the end I don't get dated trees (those with .timetree extension), I just get the follwing:

IQ-TREE report: alignment.aln.iqtree
Maximum-likelihood tree: alignment.aln.treefile
Likelihood distances: alignment.aln.mldist
Screen log file: alignment.aln.log

[alanxelena]

Dear @tothuhien

Have you had the chance to go through this issue? I have tried running the same code as before but this time instead of using a core SNP alignment as input, I used the core genome alignment that comes out of Roary (My hypothesis being that this would find less changes than an SNP analysis and therefore reduce the scale of the tree). As you will see in the figure, the dates are still off (or seem to be)

I'm attaching the i dated file generated by IQTree if it is of any help (I can't upload the alignment).
I would really appreciate some help

Many many thanks in advance.

Time_tree_core.txt

[tothuhien]

Hi, thanks for sending the tree. I open with figtree and it looks like this:

Is it correct? If so, I'm a bit concerned. The 2 groups (UEC... and the rest) seem to have large genetic distance that make the estimated times so high (the number you see are actually the elapsed dates) and it may be not correct. Are these two groups closely related or very distanced species?

[alanxelena]

Hey @tothuhien

Thank you very much for your help. They are actually all E. coli, the only difference is that they belong to different STs (The UEC group you see up there). I don't really think these differences would affect the time distances so much, do you think the opposite?

[tothuhien]

At least the tree above show that, or maybe there was something wrong. Should you check with the tree generated by iqtree before taking into account the dates? If this tree also have those branches long, then they have high genetic distances, i.e. should be something related to sequences/alignments. Otherwise, maybe something wrong with the dates.
You can obtain this tree by adding option --date-debug, then in the output folder you can find a tree file with name suffix timetree.subst

[alanxelena]

Hi @tothuhien !

Thanks for your patience and the help so far!
I've done as you say, and expectedly, these isolates as indeed distant from the rest, still, the distances are not that crazy, and actually "group" with other isolates.

Here's the file in case you want to check it.

No_time_tree.txt

Many thanks again!

[alanxelena]

Hi @tothuhien!

I feel I'm getting close to my desired output, still I'm having trouble with the dates. I've kept only the samples from the same ST and rerun the analysis using the collection date as the source of the date. This is the produced tree, you'll see that the leaves on the clusters are not that far away but the difference between clusters (in terms of age) is huuuuge.

ST_131_timetree_core_genome.txt

I would really appreciate any feedback.

Thanks thanks!

[tothuhien]

Hi,
The time tree indeed looks weird. I have looked at it a little bit and think that some given dates are not consistent with the phylogeny. Some sequences are quite similar (short distance in the no_time_tree) but having quite different dates. For example UEC31's date is 2014 and SRR11725319's date is 2020 while these two sequences are very close in the no_time_tree. Those inconsistencies introduced false signals to the dating process. Maybe there were some errors in the input dates or sequences. You can also try to exclude the outlier ones by adding option --date-outlier 3, the value 3 can be adjusted.

[alanxelena]

Hi @tothuhien,

Thank you for your response again. I'm a little confused now, I'm looking at the ST_131_timetree now and I understand what you mean that two strains are close together in the phylogeny but are apart from the time point of view. Shouldn't these variations be accounted for in the substitution rate model and related to the dates of each leaf? I mean, if a strain is extremely stable and there are very little mutations between an isolate recovered in 1900 and one recovered in 2025, the mutation rate should be taken into consideration to avoid the cases I'm having with my dataset.

[tothuhien]

Hi, sorry for the late response. This tool is supposed to estimate 1 overall rate for the input phylogeny. So if there are multiple different rates in the tree, (ex: 1 strain is very stable, the others evolve a lot), it would give inappropriate results. If just a small proportion of them do now follow the rates of the others (or could be some of them having wrong dates/sequences), you can try to exclude them as outliers as I mentioned. If possible, you can send me the data and I can have a look to detect the problem.

[alanxelena]

Hey @tothuhien !

Thanks for the reply. I understand what you mean but I'm somehow still confused about the tool behaviour since I think there will most of the times be a subgroup with different evolutionary traits. It would be great if you could take a look at the data, what would you need and where should I forward this?
Thanks again!

[tothuhien]

Can you send me the sequence alignments and the date file to tothuhien@gmail.com ?