Re: TreeAnnotator MCC Low Posterior Probabilities

[Omar Idris]

Are you constraining too much using your MRCA fossil priors try to reduce may be.

ONI

On Tue, Apr 8, 2025 at 2:54 PM Jyotishko Koley <jyot...@koley.org> wrote:

Hello all,

I've been working with BEAST for the past couple weeks to try to make a comprehensive phylogeny of 16S for a genus of frogs. There are two groups of interest to me: the mainland clade and the Caribbean clade. There are published time calibrated phylogenies for the Caribbean clade, but none with the mainland clade and certainly none that are combined. I am using sequences from GenBank as well as unpublished sequences, and I aligned using MAFFT. I have 4 MRCA priors based on fossil records and divergence dates in the literature. 

I ran BEAST with this raw alignment, which had one longest sequence for each taxa, and I combined two independent runs, one with a chain length of 20 million and one with 60 million, and my resulting posterior and prior ESS were both 28, far too low. 

I then started over, using GBlocks to trim my alignment. I ran three independent beast runs: one with 60 million chain length, and two with 30 million chain length, and still the combined ESS for likelihood and prior were both 71. All other ESS were very large, several thousand. I ran the trees file through TreeAnnotator and opened the resulting MCC in FigTree, and the vast majority of nodes had a posterior probability of almost 0 (0.001 or 0.002). I have restarted and run BEAST 3 times but I cannot get any meaningful improvement in posterior support. What do I need to change in my approach?

Thanks for any help!

--
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 visit https://groups.google.com/d/msgid/beast-users/c5d30d0c-3d21-4a1e-8af3-29ade40b10een%40googlegroups.com.