Boltz multimer with precomputed MSAs with .yaml input

[rifius]

Not sure if this should be a bug report, feature request or plain dumb mistake on my part.
Trying to run proteinfold (v2.0 or dev) multimer in boltz mode with precomputed MSAs.
I opted to specify input in a .yaml input file instead of .fasta.
multimer_test.yam contains:

version: 1
sequences:
  - protein:
    id: A
    sequence: <protein A sequence>
    msa: /full/path/to/my/protA_msa.csv
  - protein:
    id: B
    sequence: <protein B sequence>
    msa: /full/path/to/my/protB_msa.csv

When I run this (--mode boltz only) with a samplesheet.csv pointing to multimer_test.yaml, the first process being executed in the pipeline is RUN_BOLTZ, as expected, but it ends in error after successfully run extract_metrics.py. The next line in the process script is

mv "protA_protB_msa.tsv" "protA_protB_boltz_msa.tsv"

but since I am giving the pre-computed MSAs as .csv files, there is nothing to rename and the process fails.

Should this line be

mv "protA_protB_msa.tsv" "protA_protB_boltz_msa.tsv"  || true

check the input yaml if used instead of fasta, or this is not really a use case scenario ?

[tlitfin]

I think using pre-computed msas is a very valid use case but is not currently supported.