Ambiguous characters in pml

[ekopania]

Hi, I have a similar question to this one. I have an alignment with gaps ('-') and stop codons ('*'). Will optim.pml and pml ignore these by default, or will it treat them as additional states? Thanks!

[KlausVigo]

Hi @ekopania,
usually gaps are coded as ambiguous states. You want to look into the vignette
vignette("AdvancedFeatures", package="phangorn") on how to code the contrast matrix for your data.
The current models do not know about * stop codons. Be careful and check if not these character states might have been deleted. Maybe you can send me a short subset of your data so I can understand better what your data look like and provide some hints.
Kind regards,
Klaus

[ekopania]

Hi Klaus,
Thanks for your response! I will take a look at that vignette. Here is an example of one species in my fasta alignment, which includes both gaps ("-") and stop codons ("*"). Thanks!

>Mandrillus_leucophaeus
MITETAAEPTV----------------------KDAPIRTLVQRIHQLQAERAQGFRRLEEWLARVQSLR
AGRGLRVPTCRRGHRQYLRSGPDYDFARYRSTVHEVTQAFAAASREVLAVEAELAGPRSQPLLASHVRS
LQELEQTRLGTVALLQLMETPELAGQEDAVQMHQLKMKVIKTMEAISEVLQDLRFDAESAE

[KlausVigo]

Hi @ekopania,
your alignment seem not to have * :)
Currently stop codons are likely to get ignored, columns with unknown character get deleted. You would need to create your own "USER" data type. Gaps are usually treated as ambiguous states, however I have written a function gap_as_state, this transform DNA to 5 state and AA to 21 state characters, where gaps are a state on their own. This is still a bit experimental. Problem is that for amino acids rate matrices (e.g. Dayhoff, WAG, LG) do not have a rate for changes to a gap (insertion or deletion). Currently I enforce a "low" rate. For nucleotides one can estimate the rates. It is not yet well documented.

[ekopania]

Oops, it's because markdown interpreted the asterisks as italics. This is the actual sequence:

>Mandrillus_leucophaeus
MITETAAEPTV----------------------KDAPIRTLVQRIHQLQAERAQGFRRLEEWLARVQSLR
A*GRGLRVPTCRRGHRQYLRSGPDYDFARYRSTVHEVTQAFAAASREVLAVEAELAGPRSQPLLASHVRS
LQELEQTRLGTVALLQLMETPELAGQEDAVQMHQLKMKVIKTMEAISEVLQDLRFDAESAE*

Thank you for the clarification. If I created my own USER data type that allowed both "-" and "", would these both be treated as ambiguous states? Alternatively, if I just converted "" to "-" in my alignments, would this achieve the same thing?