I have one 560-bp long sequence for about 400 individuals. I understand that PhyML is faster, able to run with several model types, but non-partitioned, while RaxML can run with partitioned GTR models (not required here), the branch support is from bootstrapping only, but the result will be topologically more accurate.
That said, it’s still difficult to make a choice as I don’t know how much of topological details I will lose by running for 3 days shorter. Any advice would be very welcome.
Thank you!
Amael
Thank you very much for the suggestion, I ran some NJ trees as well (I use plug-ins in Geneious, it makes everything easier) but I was not satisfied by the results. That's why I wanted to get adequate ML trees.
I did run both PhyML and RAxML for my dataset, and the run time was about 50 h difference. In this case there were very minor variations in topology but I was wondering in general, and for future projects.
Thank you for your reply!
Amael
Along with PhyML and RAxML you ought to try TNT (See Goloboff and Catalano)
Goloboff PA, Catalano SA. TNT version 1.5, including a full implementation of phylogenetic morphometrics. Cladistics. 2016 Jun 1;32(3):221-38.
It is extremely fast and you can compare this to your other results, if they finally finish.
Scott
Thank you very much for the suggestion. I have not tried TNT so far, but I may give it a try, the abstract sells it very well.
Amael
It would be interesting to see how long the run takes with TNT besides the 50 hours in the others. Do let us know! Scott
As a suggestion, consider running smaller subsets of your data in both programs. You may start with 50 specimens. This should be fast enough for testing in a reasonable time. Compare both the time it takes to complete the analysis and the resulting topology. If there are diferences in topology export the results from one program to the other, and vice versa, to see if the numbers (ML values) come out as expected. You may expand this quick initial test in 50-specimen increments. Let us know the results.
To Scott Lyell Gardner
I ended up not running the analysis in TNT, I was finishing the project and the trees I got were adequate as it was a relatively small paper. I will see what I get from the review, and I will try TNT if needed. Thank you very much for the advice though, I will most definitively try it soon with another dataset
Thank you!
To Paulo Andreas Buckup
I hadn`t thought about doing it that way, thank you for the advice. I ran all my samples, waited quite a bit, but got extremely similar node values for both trees, with the only difference being in branch length. I will remember the method for another larger dataset coming up.
Thank you!
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running