Dear Miguel,
> This is Miguel Arenas, from the U of Vigo.
> Congrats for RAxML-NG, it is really useful!
thank you!
> We found a few technical questions.
> 1. We were wondering how the program considers indels, does it apply any model of indel evolution by
> default? I could not find any info about this in the paper neither in the manual.
RAxML-NG as well as RAxML and AFAIK other "standard" ML inference programs do not have any specific
model for indel evolution. Instead, indels are just treated as missing data (since both appear as
gaps in the MSA).
> 2. Another question is if the program can show the best likelihood per site (in addition to the
> global likelihood), we are working with protein sequences and that could be interesting.
it can print per-site log-likelihood scores computed after global parameter/brlen optimization on a
given tree:
raxml-ng --sitelh --msa ali.phy --tree best.tre --model HKY --prefix siteLH
is this what you are looking for?
> 3. Finally, we noted that if we specify an input Q matrix where an element of the matrix (a relative
> rate of change among two aminoacids) = 0 the program does not run. Maybe it is correct, but why is
> that?
Hm generally it should be allowed, although it could lead to numerical issues sometimes.
How do you specify the matrix and what error did you get?
Hope this helps,
Alexey