search

chaoszhang / ASTER

Accurate Species Tree EstimatoR: a family of optimation algorithms for species tree inference (including ASTRAL & CASTER)
GNU Affero General Public License v3.0
91 stars 10 forks source link

about wASTRAL scoring with bootstrap support #29

Open drelo opened 3 months ago

drelo commented 3 months ago

Following your Q&A I am trying to use Weighted ASTRAL - Hybrid (i.e. astral-hybrid) with the bootstrap gathered for my 3000 gene trees.

Q: I have gene trees with branch lengths and Bootstrap/Baysian supports and I know that horizontal gene transfers and hybridizations are rare. A: I recommend Weighted ASTRAL. It utilizes branch lengths and supports to improve accuracy.

IQ-TREE provides the consensus tree of each bootstrap so I used file with all the individual gene trees with the consensus tree or support values concatenated (3000 lines, 3000 trees)

Example of the first line in this file

(Cas:0.0555076248,((Cra:0.0152295646,Bot:0.0021838675)96:0.0067535899,Lon:0.0190199014)52:0.0015067178,((Mer:0.0028333550,((Ard:0.0014141248,Spe:0.0014208224)49:0.0000023700,Ste:0.0014124767)43:0.0000023700)55:0.0002869319,Oas:0.0064360194)70:0.0129822513);

then I run...

./ASTER/bin/astral-hybrid -C ./GENES/contrees3991.trees

I have two questions then. The annotated support is the quartet support right? I guess the calculations included the support value in each tree (bootstrap values) for estimating this QS then.

Thanks

Andrés

chaoszhang commented 3 months ago

I am a little bit confused. I hope this will answer your questions.

Regarding IQ-TREE, your input should be a concatenation of .treefile files. I assume those are ML trees instead of consensus of bootstrap trees. Regardless, your example first line looks fine.

Regarding your command, I would run ./ASTER/bin/astral-hybrid ./GENES/contrees3991.trees or ./ASTER/bin/astral-hybrid -i ./GENES/contrees3991.trees.

Regarding your questions, those suports are local posterier probabilities (https://academic.oup.com/mbe/article/33/7/1654/2579300) and yes they integrated the support values. You can also run normal astral ./ASTER/bin/astral4 ./GENES/contrees3991.trees and you should get a slightly different result because normal astral does not consider weights.

I am traveling. Excuse for typo.

drelo commented 3 months ago

Thanks for your help. I think I tackled my initial confusion. I have multiple trees but from multiple loci. So in an analogous fashion to Astral-III I thought one would provide 1 bootstrap consensus tree per loci (that version had the -b flag and one provided the path of each file with each bootstrap trees for each loci). Thanks for the clarification on local posterior probabilities I read about it after I posted here, I think now I can estimate them for a collection of the best trees from each loci, like you said -i ./GENES/contrees3000.trees

chaoszhang commented 3 months ago

I see. Weighted ASTRAL has not yet supported -b option. One thing I would do is to first compute a species tree using wASTRAL, and then use ASTRAL-III -q option to score such tree and use -b option to compute bootstraps at the same time.