Stockholm input format

[vagkaratzas]

Hi! Is there any easy way to make pytrimal work with stockholm MSA inputs?

Thanks!

[althonos]

Hi @vagkaratzas !

Stockholm is not supported by trimAl directly in the original C++ codebase, but their code is generic enough that one could add support for reading Stockholm files. I could try to implement it but I can't exactly say when that will happen.

In the meantime, you can always use a third-party library such as Biopython to load the alignments, and convert them to a pytrimal.Alignment object with the Python API, e.g:

import Bio.AlignIO
import pytrimal

bio_ali = Bio.AlignIO.read("example.sto", "stockholm")
pytrimal_ali = pytrimal.Alignment(
    names=[record.id.encode() for record in bio_ali],
    sequence=[bytes(record.seq) for record in bio_ali]
)

[vagkaratzas]

This was already extremely helpful, thanks ;)

[vagkaratzas]

And I guess neither can it work with pyhmmer DigitalMSA objects, correct? Even though I:

msa = msa.textize()
pytrimal.Alignment(msa.names, [seq.sequence for seq in msa.sequences])

I get this error: ValueError: Sequence length mismatch in sequence 5: 90 != 109, probably because the gaps don't show in the sequences?

[althonos]

Indeed TextMSA.sequence gives you the constitutive, unaligned sequences, but you can use TextMSA.alignment (agreed this is not very clear) :

pytrimal.Alignment(msa.names, msa.alignment)

[althonos]

This makes me realize that while the TextMSA constructor expects gapped sequences, TextMSA.sequences gives you the ungapped ones... That's not super straightforward, maybe I'll refactor that in pyhmmer.