Functional clustering of proteins

Question

[nadavi]

Hey everybody,

I have a list of N proteins off NCBI's NR database, where the info I have for each protein is its name and sequence.

Is there a way I could cluster said proteins by their molecular functionality?

There's no point in re-inventing the wheel everytime, but I didn't find anything when looking for articles where people have done something similar, so any help would be appreciated.

Thanks!

[Meisam]

If you only have a list of names (no enrichment or expression scores to compare) you can simply do an over-representation analysis with a tool like gProfiler or DAVID. DAVID is normally not recommended anymore as its database is outdated, but it's straightforward enough to worth give it a try.

[Jean-Karim Heriche]

Just to point out that while overrepresentation analysis can be useful to analyse a list, the question is about clustering which isn't the same thing.

Answer 1

[Jean-Karim Heriche]

Since most clustering algorithms operate from a pariwise distance or similarity matrix, you need to compute one such matrix that captures relevant similarities between the proteins in your list. Molecular functionality is somewhat imprecise. I generally take it to mean what's covered by the molecular function domain of the Gene Ontology which you can use to compute some similarity between proteins, for example with the R Bioconductor package GOSemSim.

[nadavi]

Yeah you nailed it with what I am attempting to do, I'll give it a shot.

Thanks a lot!