Grouping protein domains into categories?

Question

[exin]

I want to compare the domain composition of a few of my cell samples.

After mapping peptide sequences on Pfam, I have thousands of domain hits across my dataset. Comparing across my samples hardly makes sense.

Are there any references / databases for domain categories? Similar to the clans/ superfamilies in Pfam but something broader, e.g. kinases, zinc fingers, ankyrins... ?

[Joe]

To my knowledge, no - I don't think it gets much broader than "Zinc-finger containing protein" if you're specifically interested in the domains themselves.

The next level up would be pathway ontologies through Kegg/GO etc, but its not entirely the same thing.

[exin]

Thank you!!!
So no way to categorise them other than manually going through the domain names? - Currently they're like zf-C3HC4, zf-C2H2, zf-RING2....

[Joe]

AFAIK, no; but I won’t claim to be an expert here.

If your dataset is reasonably well organised/curated you may be able to get some of the way with assorted regex magic, but that’s rarely the case with annotations.

Answer 1

[jgreener]

There are a few resources for organising protein domains:

• CATH sorts domains by structure into class, architecture, topology and homologous superfamily.
• ECOD sorts domains by evolutionary relationship into architecture, possible homology, homology, topology and family level.
• SCOP and its later incarnations, SCOP2 and SCOPe, also sort domains by structure.

I would recommend the first ECOD paper as a good read for how to organise protein domains and how these resources relate to each other.

Hopefully one of those resources will help you make sense of your domains.

[exin]

Great!! Thank you for these!!