v1.1.0 PC_V1 scores are all 0.5

[NatJWalker-Hale]

Hi @CarineRey,

I have just upgraded to v1.1.0 via docker, and I've noticed that in my tested runs, the PC_V1 scores in the output are all 0.5, which is presumably in error. I am using a transition scenario that excludes some descendants of a transition node, so perhaps that is causing it, but I did not observe this error on v1.0.1.

I've attached the tree, alignment and scenario file necessary to reproduce the run. The command was pcoc_det.py -t tree.nwk.tx -aa seqs.fa.txt -m $(cat scenario_exclude.txt) -est_profiles C60 --gamma --max_gaps_allowed 10 --max_gaps_allowed_in_conv_leaves 10

Thanks again for your help and the great tool,

Nathanael
scenario_exclude.txt
seqs.fa.txt
tree.nwk.txt

[NatJWalker-Hale]

Can confirm that it is not the scenario excluding some descendant tips causing the issue. Would be great to know if the PCOC_V1 scores are still valid in this case?

[NatJWalker-Hale]

A further update to this - when checking the infos output from bppml, it seems the likelihoods are invariant to the second profile selected - i.e. the log likelihoods in 10_47 are the same as the likelihoods in 10_48. I've looked into estim_data.py and it doesn't seem like there is any obvious error - so perhaps this is a bug with bppsuite v3?