exact source of the HMMs

[jvollme]

Hi, I am trying to combine marker sets derived from different sources. So I am wondering: How exactly were the underlying hidden markov models derived for fetchMG?

Were the raw protein sequences downloaded from the cog database and aligned by yourself? If yes, could you describe the aligner that was used as well as the alignment parameters? Would it make sense to update the alignments and models based on the 2015 and (apparently) 2020 updates to the COG database?

Or are the models and/or alignments of these cogs freely available (could only find the raw protein sequences on ncbi)? And if yes, then again, would it make sense to update these fetchMG-models based on the 2015 and/or 2020 versions of the cog database?