augur/clades.py: change clade assignment from mutations to sequences …

[rneher]

…to avoid unassigned nodes when reference sequence with respect to which mutations are defined is not part of the build. This still needs to be checked with vcf (cc @emmahodcroft ). there are also now obsolete parts that could be deleted (such as get_node_alleles)

[trvrb]

Thanks for this @rneher. I had thought that it wasn't safe to assume that the root node would be populated with a full nucleotide sequence and aa_sequences. But I see you can also input these separate as --reference. I have to admit that I'm now a bit confused about all the different ways we're passing around root sequence vs reference sequence vs JSON vs FASTA. I can try to gain clarity.

[rneher]

what I had pushed earlier worked for flu, but would have broken vcf workflows. the latest commit is not going to work, but is my attempt to unify them. I'll have to test and fix this tomorrow (somehow the ssh access to the cluster doesn't work).

[rneher]

My current thinking was to implement something that

• without a reference/root sequence works (or not) as before
• if the nt-muts/aa-muts jsons have sequences, it will use those to disambiguate states that not diverse
• if a reference is passed in for vcf, it'll use those instead. this is not yet implemented
  This can all be achieved by re-interpreting the mutation dictionaries as sequence look-ups that fall back to the reference if no mutation has been observed at a particular position.

[rneher]

this should work now.