Bioinformatics - basic homology to attempt to better understand protein function

[Jen_Scott]

Hi,

Sorry to bother you all. I’m not a bioinformatician, but I happen to have a need to do some bioinformatics, and I’m getting a bit lost trying to teach myself so could do with some advice.

I have a protein sequence for a 7TM-GCPR, which I’m trying to find the most likely function for, specifically the ligand. I’ve pBLASTed it and it comes back with a number of hits, the problem is that most of the hits are putative proteins for a particular ligand, but they have not yet been confirmed empirically . I tried PSI-BLAST, but unfortunately as the vast majority are only putative, it didn’t really give me much to go on!

Within the results are 5 proteins that have E-values ranging between -28 to -39, these 5 proteins have been characterized and have the same known function (i.e. same ligand), however the identities are all around 28% and the similarities are all around 48%. On the basis that I know what these 5 proteins do, I want to see how similar my protein sequence is to these 5 protein sequences. What I’ve done is build a multiple sequence alignment, which I’ve done using Clustal. I then copied the MSA and pasted it into HMMERsearch and submitted it against the RefSeq database, and it came back with a much better E-value for my protein, E-62 (Same identities, but 64% similarities).

I fully understand that homology does not necessarily mean that it will have the same function. But seen as nobody appears to know anything about this protein, it's the best I can go on!

So I guess my question is, does that sound like the right course of action to you to try and compare my protein to these 5 sequences? Or am I completely off the mark?

Many thanks in advance

[GenoMax]

Jen: If you are interested in trying to find a "function" for your protein then a better option would be to start your blast using the protein data bank or the Swissprot database (at NCBI blast) to search against. Entries in these two DB's are good quality annotated/curated proteins. If you are able to find a hit in the PDB database that can give you a jump start on trying to model your protein using structure prediction meta servers (http://www.expasy.org/proteomics/protein_structure). (Hint: You should find something in PDB is your protein is indeed a GPCR based on the name you have used).

You can also try the SIB-Blast+ server at Expasy: http://web.expasy.org/cgi-bin/blast+/BLAST_new.pl.