RAxML-NG bootstrap files

[Sonia]

Hello. I am generating trees in RAxML-NG in Cipres. I am using the "all-in-one" analysis mode which I understand draws support values using both Felsenstein’s Bootstrap (FB) and Transfer Bootstrap Expectation(TBE) on the best-scoring tree. It seems that the FB file is included as "label" when I download the infile.txt.raxml.support file and open it with FigTree but I do not know where the TBE values are stored and how to visualize them on the best ML tree. Thanks for your help.

[Mark Miller]

Hi Sonia, thanks for the question.

Can you please send me the file _JOBINFO.TXT for the job in question?

Mark

[Mark Miller]

Hi Sonia,

First I am by no means a raxml-ng expert. So we might need help from the raxml mail list.

But here is what I see:

Your command line is

raxml-ng_1.0.1_expanse -msa infile.txt --workers 10 --bs-trees autoMRE{1000} --all --model part.txt --tree rand{500} --tree pars{500}

But the manual says the TBE file must be requested by a specific command flag, which your command line lacks:

https://github.com/amkozlov/raxml-ng/wiki/Tutorial

Alternatively, we can compute so-called Transfer Bootstrap Expectation support metric (Lemoine et al. 2018), which is supposedly more appropriate for very large trees:

raxml-ng --support --tree T15.raxml.bestTree --bs-trees allbootstraps --prefix T15 --threads 2 --bs-metric tbe

Which produces a file: T15.raxml.supportTBE, that I presume is the one you need?

The CIPRES interface offers this option:

Specify the branch support metric (bs-metric).

If you chose this option, and the command line didn't form, let me know and I can explore what went wrong.

Otherwise, maybe you can clone and resubmit, and make sure that option is chosen.

That is the limit of my knowledge right now. Let me know if I can help further.

Best,

Mark

[Mark Miller]

Also, you have the option of selecting "Both" which will add the flag. --bs-metric fbp,tbe

[Sonia]

Hi Mark. Thank you so much for your help. The "Using RAxML-NG in Practice" tutorial states that "all in one" in RAxML-NG offers an analysis mode that will do ML inferences on the original MSA, infer bootstrap replicate trees, and "draw support values using both FBP and TBE on the best-scoring tree with just a single command". It didn’t occur to me that the same in CIPRES may involve a different command - I should have checked, sorry. My runs have used an awful number of CPU hours and that observation also fed into my assumption that the TBE calculations must have been included in those runs. Anyway, many thanks for clarifying and providing solutions. I'll follow your recommendations!

[Mark Miller]

Hi Sonia,

It is possible that I am completely incorrect on this. I think you should ask the raxml google group. The command for cipres should be identical to those for the code itself.

Sorry if I misled you.

Mark

[Sonia]

Hi Mark. Thanks so much. You did not misled me. My raxml google group post also got a similar response as you provided, that when using raxml-ng on CIPRES this choice must be made separately.