Pre-computed alignments?

[Sigve Nakken]

Hi,

Thanks for a neat tool! I see you have attached a few existing protein sequence alignments with your package. Are you aware of any resource/database in which you can fetch such data (i.e. pre-computed alignments) for any human gene? Or do you need to compute this for any gene of interest using existing multiple sequence alignment tools?

kind regards,
Sigve

[Ramiro Magno]

Hi Sigve,

Thanks for reaching out!

Just like you said, we do have a few alignments bundled with {agvgd}. These alignments are the ones also provided in agvgd's original website.

As you might know, a multiple sequence alignment is something of an art. :) I don't know of any resource that offers such pre-computed alignments for any given human gene. In Ensembl you can download an alignment of orthologues which is computed on-the-fly for you. But it is not something you can do in batch-mode, I reckon.

For example, for TP53, we may go here: https://www.ensembl.org/Homo_sapiens/Gene/Compara_Ortholog?db=core;g=ENSG00000141510;r=17:7661779-7687538, then hit Download Orthologues. Then choose Alignment -- amino acids.

There is also something similar, which is the pairwise alignments between an orthologous sequences and a human sequence. You can retrieve them using Ensembl's REST API: https://rest.ensembl.org/documentation/info/homology_symbol. In the experimental package {protean} we are using this method to create what we call sequence profiles. Sequence profiles are built out of those pairwise alignments, but we remove gaps from the human sequence (used as a reference), and create a pseudo-alignment of those orthologous sequences. Like I said, the package is experimental, but you are welcome to try it, if you find it useful. The function to get those profiles is: get_profile().

But, in general, you need indeed to compute your own alignment, and probably curate it manually. The original authors of agvgd did this for a few genes (the ones we include with {agvgd}), but I think it is not easy to perform such task automatically and genome-wide, at least, not without questioning the quality of the alignments.

cheers, RM

[Sigve Nakken]

Thanks a lot, Ramiro, for your swift response! This is indeed very helpful and useful information. I was actually suspecting that making a multiple sequence alignment is not straight-forward, considering which species to include etc. I'll pursue this further, and see if I am able to make something work out :-)

cheers,
Sigve