Weird trace, mirror images

[Mark DiMeo]

Dear all,

I was trying out *BEAST with a 3 mitochondrial and 4 nuclear gene data set when I ran into this issue in my tree likelihoods for two of my mitochondrial genes. Everything else seems to have good ESS values but those two were low at 48. Looking at their trace files they both have 2 spikes in an otherwise good looking plot. The weird thing is they are mirror images of each other as you can see in the attached pdfs. I'm concerned that they might be messing up my results but I can't figure out how to fix it as it's persistent. 

Should I be worried? Is it an issue when my other parameters look fine? Any help would be appreciated!

Cheers,

Mark 

[Huw A. Ogilvie]

Hi Mark,

Have you linked the mitochondrial gene trees? Usually we assume no recombination within mitochondria, so the trees for all mitochondrial loci are linked.

Of course as with anything biological this assumption does get violated from time to time, so perhaps if you have linked the trees and there has been recombination the MCMC is switching between a gene tree that accomodates both loci, and a gene tree that's a better fit for the "WeirdTrace2" locus and a worse fit for the "WeirdTrace" locus.

[Mark DiMeo]

Yes I did link them. I guess it could have some recombination but I am aware that is very rare.

[Huw A. Ogilvie]

I would suggest unlinking them and comparing the posterior distributions of gene trees for the different mitochondrial loci as inferred by *BEAST. Then if they are radically different, there could be a number of reasons why that could be the case, e.g. as discussed in this review https://link.springer.com/article/10.1134/S0026893313010056