bayesian and ML method -similar output

125 views
Skip to first unread message

B Jadeja

unread,
Mar 14, 2021, 2:46:45 PM3/14/21
to beast...@googlegroups.com
Dear All, 

Has anyone compared their phylogenetic trees derived from Bayesian and Maximum likelihood method? Is it possible to have a similar tree pattern from both these methods? I have been using some viral genomes with BEAST and RaxMl (applied the same model GTRGamma in both methods) and the best tree from both the strategies is exactly the same. Any insights/comments are highly appreciated.
thanks

best 
Niti

Omar Mejía

unread,
Mar 14, 2021, 2:56:13 PM3/14/21
to beast...@googlegroups.com
Theorically, the most parsimonous, the ML and the bayesian tree must be identical if you we're able to recover the "true" phylogeny

Regards

--
You received this message because you are subscribed to the Google Groups "beast-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beast-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/beast-users/CAEpwqv57x2mtg4OLM5rSAP-xTDDH3dDou9AN_9fYSyzgXOtOqw%40mail.gmail.com.

Alexei Drummond

unread,
Mar 14, 2021, 4:35:55 PM3/14/21
to beast-users
Dear Omar,

I am sorry but your statement is incorrect. The Felsenstein zone was first described in 1978 and shows that for some true phylogenies, maximum parsimony will positively mislead, whereas statistically sound methods such as ML and Bayesian may be able to recover the true phylogeny with the same data.

Please see this seminal paper:

Felsenstein J. 1978. Cases in which parsimony or compatibility methods will be positively misleading. Syst Zool. 27(4):401–410

Cheers
Alexei
> To view this discussion on the web visit https://groups.google.com/d/msgid/beast-users/CAOEgSgmAiNhuC57YPwcmzyPmJ7XhJ6OKnqCzFoEuFGVHFpteCw%40mail.gmail.com.

Alexei Drummond

unread,
Mar 14, 2021, 4:39:08 PM3/14/21
to beast-users
Dear Niti,

Bayesian and ML are both good statistical methods. Bayesian methods can be shown to be superior in certain circumstances, but I would not be surprised that in particular instances they produce the same tree topology, especially if the phylogenetic model chosen in each case is similar. It is important to understand that Bayesian methods have a tree prior, whereas ML methods do not. So in general they won’t produce the same tree since they use different information and make different assumptions.

Cheers
Alexei

Niti Jadeja

unread,
Mar 14, 2021, 11:44:54 PM3/14/21
to beast...@googlegroups.com
Dear Alexei & Omar, 

Thanks for your comments & explanation, my sequences are sampled very close in time, I am hoping to have more genomes from samples more distant in time in this year; drawing similar comparisons (ML & BS) in future will be very interesting. Your replies were really timely and helpful.
Thanks again,
Niti

From: beast...@googlegroups.com <beast...@googlegroups.com> on behalf of Alexei Drummond <alexei....@gmail.com>
Sent: Monday, March 15, 2021 2:09 AM
To: beast-users
Subject: Re: bayesian and ML method -similar output
 
Reply all
Reply to author
Forward
0 new messages