search

soedinglab / plass

sensitive and precise assembly of short sequencing reads
https://plass.mmseqs.com
GNU General Public License v3.0
149 stars 14 forks source link

Protein sequence abundance #13

Open cmorganl opened 5 years ago

cmorganl commented 5 years ago

Hi there,

I've just started using PLASS, specifically plass assemble, and I really like it!

This isn't an issue but a potential enhancement. I was wondering how I might be able to recover abundance information for each protein assembled. Its not as simple as with nucleotide contigs since I'm unable to align the reads back to the assembly in this case.

Could the abundance of proteins be included in the header, as SPAdes does? Alternatively, a table mapping header to its respective abundance would be convenient as well. I hope I haven't misinterpreted the output and its already provided :)

Thanks! Connor

martin-steinegger commented 5 years ago

Yes I would love to have this as well. 👍 But so far I did not come up with an solution how I could gather this information.

It is hard to set rules what alignments should be part of the abundance computation since plass allows to assemble similar proteins and not just exact once.

We estimated the abundance in the plass publication using the mmseqs map workflow. This workflow performs a six frame translated search and has strict mapping thresholds.

apcamargo commented 5 years ago

Wouldn't it possible to use an EM algorithm to "distribute" the read among its multiple hits?

martin-steinegger commented 5 years ago

@apcamargo yes this a good idea. But do you know if EM would be fast enough to handle this large amount of data?

apcamargo commented 5 years ago

I don't know. It would depend on the total amount of proteins and reads. I imagine it would be much slower than a typical RNA-Seq quantification, but not unfeasible. But that's just my impression, I've never done something like that.

mooreryan commented 5 years ago

I was just thinking about this as well...are there any current plans to include this in plass, or would it be better for now to try the mmseqs map workflow?