[Feature Request] Allow multiple tree constraints

[HS6986]

Dear RAxML-NG Developers,

I'm running RAxML-NG v1.2.2 with topological constraints and I've realized that RAxML-NG doesn't allow multiple tree constraints; it recognizes only the first tree that appears in a constraint file. It may sound strange to set constraints by multiple trees, but this is because I want to set floating taxa (taxa not included in a constraint) separately for each constraint. This is not a problem in MrBayes, TNT, or TreeSearch because they provide features to set floating taxa separately for each constraint. IQ-TREE shows the same behavior as RAxML-NG (reported in iqtree/iqtree2#408).

Here's the folder of my RAxML-NG reanalysis of a morphological dataset from Shao et al. (2020) with constraints, showing that RAxML-NG doesn't consider multiple tree constraints.

This feature is needed for my research. I would greatly appreciate it if you could implement it in a future RAxML-NG release, if possible.

[HS6986]

Dear RAxML-NG Developers,

I received a reply from an IQ-TREE developer on this matter (iqtree/iqtree2#408), but he said that currently he does not know how to implement this feature in IQ-TREE and is facing difficulties. If you could suggest a strategy to implement this in a maximum likelihood framework, it would be greatly appreciated.

Thank you for your time and support.

[HS6986]

Dear RAxML-NG Developers @amkozlov,

Hi, just following up — this discussion has been open for several months without any response.

I'd really appreciate it if you could take a look when you have a chance.

Thanks for your attention.

[amkozlov]

Sorry for the late response.

As Minh already noted in the IQTree repo, multi-tree constraint implementation would not be straightforward. In fact, given my experience with debugging the constrained tree search with free-floating taxa (=partial constraint), I would be rather reluctant to implement something even more complicated. The overhead for maintenance and covering the long tail of weird edge cases could be way too high.

Are you sure you can't get away with combining all constraint trees into a single one?
How do you specify your constraint for TNT/MrBayes?