MERGE PARTITIONS WEIGHTS-CRITERIA

[Alexandros Vasilikopoulos]

Dear iqtree developers,

I want to ask what is the criterion for calculating the similarity of subset when using the
TESTMERGEONLY and MF+MERGE options. I know that in partitionfinder there is an option called weights 'x,x,x,x'
which basically correspond to weight values for model, rate, frequency, and alpha parameter of the gamma distribution.

Is there sth similar in iqtree and if not how is the similarity calculated?

Many thanks
Alex


[Bui Quang Minh]

Dear Alex,

IQ-TREE only takes the rate into account, i.e. model, frequency or alpha are disregarded. In other words, the similarity between partitions is based on the rates alone. I believe that is also the default option in PartitionFinder. At some point I also thought about supporting such a weighting. But then Rob Lanfear told me that users rarely change the default option. Moreover, there is no good sense of how to weight the parameters.

Therefore I decided not to implement such option in IQ-TREE. If you however have some compelling reason then I’m happy to reconsider it.

Cheers, Minh

--
You received this message because you are subscribed to the Google Groups "IQ-TREE" group.
To unsubscribe from this group and stop receiving emails from it, send an email to iqtree+un...@googlegroups.com.
To post to this group, send email to iqt...@googlegroups.com.
Visit this group at https://groups.google.com/group/iqtree.
For more options, visit https://groups.google.com/d/optout.

--

Bui Quang Minh

Center for Integrative Bioinformatics Vienna (CIBIV)

Campus Vienna Biocenter 5, VBC5, Ebene 1

A-1030 Vienna, Austria

Phone: ++43 1 4277 74326

Email: minh.bui (AT) univie.ac.at

[Alexandros Vasilikopoulos]

Dear Minh,

Thanks a lot for the answer. Could please elaborate a bit more about this rate calculation? Is this a substitution rate or an evolutionary rate?
And if the second how is it calculated?

Best wishes and thanks again
Alex

[Bui Quang Minh]

Hi Alex,

With rates I mean the sum of branch lengths of the partition tree (or tree length). Fast/slow evolving partitions will have high/low tree length. ModelFinder will consider to merge partitions having similar evolutionary rate. You can think of this partition rate as an average rate across all sites in the partition. This is of course not substitution rates (A-C,A-G etc.). 

Note that this discussion is only valid when you use -rcluster option.

Cheers, Minh