search

matsengrp / cft

Clonal family tree
5 stars 3 forks source link

smarter selection of parsimony tree #176

Closed wsdewitt closed 5 years ago

wsdewitt commented 7 years ago

dnapars generally outputs severally degenerately optimal trees. The current process_asr.py script parses the dnapars output, but only takes the first tree it finds. We have ideas for resolving parsimony degeneracy (like with motif likelihoods), so process_asr.py needs to be updated to parse the full list of trees, and select one (initially just the arbitrary first one, but later something smarter like motif ML tree).

metasoarous commented 7 years ago

Great! If you could have it add an output for a newick file containing all of the trees, that would be wonderful as well. We can use this downstream to compute support for the splits.

metasoarous commented 7 years ago

@WSDeWitt What's the status on this?

wsdewitt commented 7 years ago

I have an oldish issue branch where I'm parsing all the trees, but still only returning the first one arbitrarily. I'm happy to work toward merging it, especially if we're getting to a place with @dawahs's motify code that we could rank them.

matsen commented 7 years ago

May I upvote this issue? It's really the next thing to try for getting more accurate point estimates.

@metasoarous , if you can work together with @dawahs to get his motif-model stuff hooked up with what @WSDeWitt has been doing, that would be magic ✨ . If we can make it somewhat stand-alone then @krdav can test it in his same framework.

dawahs commented 7 years ago

The relevant modules are here:

https://github.com/matsengrp/samm/blob/master/cft_tools.py

@metasoarous let me know if you want to meet up/chat to go over this.

metasoarous commented 5 years ago

I'm not sure what happened on this issue, but I think it was some combination of deciding with Laura that the ML trees were doing what we wanted them to (after fixing some tree viz quirkiness with the old ete3 rendering system), and other work taking priority. In any case, just checking in here to see if this is something which we want to resurrect, or whether we want to focus on custom phylogenetic methods, or seeing if we can't get some other tree program to do what we want it to.

@matsen Thoughts on this?

matsen commented 5 years ago

We don't use parsimony for anything any more, right? If not, I think we can close.

metasoarous commented 5 years ago

Correct, we are not.