Unexpected behavior in IQ-TREE 3.0.1

[mmatschiner]

Dear developers,

Thanks for developing a great tool!

I thought I'd report some surprising behavior of IQ-TREE 3.0.1 that might not be on purpose:

• With "-t INPUT.tre" as the only option, IQ-TREE 3.0.1 runs the PD algorithm on the input tree, and it also does so when "--scfl 100" and/or "--prefix OUTPUT" are used together with "-t INPUT.tre". Should the PD algorithm perhaps be called explicitly with a separate option?
• With "-t INPUT.tre -s INPUT.nex --scfl 100", IQ-TREE 3.0.1 runs a tree search before computing site-concordance factors with likelihood. So the topology of the output tree differs (in my case) from that of INPUT.tre. Shouldn't the tree provided with "-t" be the "reference tree to assign concordance factor" (help text for 3.0.1)?

With best wishes,
Michael

[aldendirks]

I was dealing with the issue of a different tree after running -scfl. The IQ-TREE concordance tutorial talks about using -te and -blfix to use the same topology and branch lengths. -t gives a starting tree, but it is modified (optimized) during the scfl likelihood process, resulting in a different output. This does not happen with gCF because that only compares loci tree to the input tree. This info is not available in the -h output. I'd recommend putting it there, too.