gtr.rates and frequencies for a single partition flatline in independent analyses

[David Lohman]

Greetings, 

I'm running a BEAST dating analysis with 197 taxa and 100 loci in 16 partitions. The first partition includes a single locus--the only mitochondrial locus in the dataset. The other 15 partitions are nuclear protein-coding loci with variable numbers of loci grouped by ModelFinder. Stepping-stone analyses selected a Yule model for the tree and 2 clocks: one for the mitochondrial locus and a second for the nuclear loci. Both are uncorrelated relaxed clocks with a lognormal distribution. There are 3 secondary calibration points with normally distributed priors. All other priors use default values.

In 2 independent analyses, the 6 gtr.rates and 4 frequences for the mitochondrial locus (only) stopped sampling midway through the 250 million generations, making the traces flatline. See attached images.

This post from nearly 8 years ago reports a similar problem, but there are no replies:

https://groups.google.com/g/beast-users/c/qHEVXX3qd74/m/X7W_LJiIBAAJ

I have run many similar analyses on comparable datasets and never encountered anything like this before.

Any advice on how to fix this problem would be greatly appreciated!

Best wishes,

David Lohman