analysis time

[chris blair]

Hi all, 

I am performing some raxml-ng runs and am noticing that it takes much longer than IQ-TREE, even with model selection incorporated into the latter. Right now I am running a default ML search with 10 random and 10 MP starting trees and it looks like it will take about 12 hours to finish using 48 threads (cores). Subsequent bootstrapping will take much much longer. I am not receiving any warning or error messages, and my goal is to compare the best topologies inferred between programs. The matrix consists of 82 taxa and 7,543,227 sites. I am using a GTR+G substitution model. Curious to hear your thoughts.

Chris

[chris blair]

Colleagues, 

Apologies if this is a duplicated message as I am not sure that it posted the first time. I am running a raxml-ng analysis to compare with IQ-TREE. I notice that raxml-ng is running much slower than IQ-TREE, even with the latter first inferring the best substitution model. Right now I am running a default ML search (10 random and 10 MP starting trees) and this alone will take about 12 hours for raxml-ng. Bootstrapping would obviously take much longer and may in fact be prohibitive with this data set. The matrix consists of 82 taxa and 7,543,227 sites and I am using 48 threads. Again, IQ-TREE used far fewer threads and was much faster, which promoted these questions. Any thoughts would be appreciated. 

Best wishes,

 Chris 

[Oleksiy Kozlov]

Dear Chris,

could you please attach log files from both raxml-ng and iq-tree?

In general, 20 starting trees are likely too much for such an "easy" alignment.
We are now switching to the "adaptive" search strategy which will automatically choose the number of
starting trees based on the alignment difficulty:

https://academic.oup.com/mbe/article/40/10/msad227/7296053

https://github.com/amkozlov/raxml-ng/wiki/Installation#building-development-branch

Best,
Oleksiy

> --
> You received this message because you are subscribed to the Google Groups "raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to
> raxml+un...@googlegroups.com <mailto:raxml+un...@googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/msgid/raxml/53a83c8e-d835-4eb2-
> a1cf-2058559fc4adn%40googlegroups.com <https://groups.google.com/d/msgid/raxml/53a83c8e-d835-4eb2-
> a1cf-2058559fc4adn%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Christopher Blair]

Dear Oleksiy, 

Thanks for the response and the new updates to the program. Please find attached log files for both raxml-ng and IQ-TREE. Please note that I kept receiving a numerical underflow warning with IQ-TREE. I brought this up on their discussion board and was told that the results are reliable as long as a final likelihood was calculated. 

 MVZ228761_ornatus_pruned.phy.log

Chris

You received this message because you are subscribed to a topic in the Google Groups "raxml" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/raxml/YJ7688In7nQ/unsubscribe.
To unsubscribe from this group and all its topics, send an email to raxml+un...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/raxml/68eed4ea-7c9f-4ffa-a34a-1ff79e555b58%40gmail.com.

--

<{===================~~~~~~~~~~~
          /\                      /\
Christopher Blair, Ph.D. 
Executive Officer, Biology PhD Program

Professor

Graduate Center and

New York City College of Technology 
The City University of New York
CBl...@citytech.cuny.edu; cbl...@gc.cuny.edu
Website: https://sites.google.com/site/christopherblairphd/home
Office: Academic Complex 501D; Ph: 718-260-5342 (City Tech)
Lab: Academic Complex 401 (City Tech)