Hi,

When you BLAST against multiple databases it is expected that you would get different IDs for same sequence. Did you try ID mapping between different databases? tht would take care of multiple databases. secondly in your homolog search of unnamed proteins if you use one organism at a time as filter (ex: mouse homolog) should help you.

-

Rachita

Dear Hajer,

First check the lenght of your assembled protein sequences. The smaller they are, the less informative would the blast be. Once you have considered that, I assume you are working with a non model organism, right? If that is the case, use databases such as RefSeq and restrict the species you want to blast against: use the one that is most closely related to the species you are working with.

As for which parameter you should trust the most, I'd go first with the E-value but for each case consider also the coverage: if your best hit has a good e-value but a poor coverage, then you can discard it. I would also recommend that you use a bidirectional best hit. Once you blast your query protein, select the best hit and blast it against a local database including all your protein IDs, which ideally will retrieve your original query as the best hit.

A final note: have you tried to look for protein domains instead? Using maybe superfam or any other protein domains databases. It may worth it depending on what analysis you want to conduct hereafter.

Best regards.

Use BLAST2Go, you will get all the multiple IDs in one column. It will even give the search results based on the presence of domains in your protein, i.e, InterproScan in BLAST2Go.

The tool is very easy to run and install in windows/linux. So long you don't have a interfering internet connect.

Thanks to all of you guys.

Rachita  could you elaborate more how when I BLAST against multiple databases it is expected to get different IDs for same sequence . it supposed to give a range of  matching , Right? as from my short experience on  blasting, my list of hits  that match the sequence may vary , long or short, or no matching at all ( as I get for some ). I understand that one sequence supposed to be a homologous of  one protein at least from same family and SAME or at least simillar result I supposed to get on all DP.

how to do  please ID mapping between different databases, have no idea ?

Also I tried first to find a a homologous  protein in my organism database , but  although my organism is a model organism due  to lack of annotation I did not get any match. Hence, I tried to blast against the closest species, which is a model organism, and if I still did not succeed in my homolog search I go to a higher classification ,where my organism belongs to  like the Phylum. The problem here all of that did not  help  with those unnammed proteins , as  my blast resulted in proteins from   totally different family!! so  how I decide which one of these  my protein sequence is belongs to!

Dear Gustavo  

 the length of my  protein sequences is between 20-800  AA. Unfortunately, most of them are small, but they are valuable to me, so need to identify. Also  I am working with a  model organism , but it is not fully annotated, so did not get any result on its DB neither on the most closely related species which forced me  to wide my search blast to get a hit. I did also restrict the search  on some DP that offer this option , but some that give a match do not  give the restriction option .

As for the parameters you mentioned,  definitely I look at the E-value, but some DP that did give me a result did not show the E- value , most importantly. Also, as most of my AA sequences are short (20-150), so trusting a high coverage  looks not possible with such short AA, and same challenge for looking for  the protein domains.  Is it enough to trust only the E- value if that was my only option?  

I googled the bidirectional hit and how to use it, but could not get much help , as it is new to me. Could you send me the link please. with many thanks

Different databases will have different accessions for the same protein.

Hey Salem, you can consider evalue and bitscore for infereing homology. Here is paper would help you 

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3820096/

If you have short proteins, you should use a similarity scoring matrix like PAM30 or BLOSUM80 with BLAST.  Or you should use a program designed for peptides, such as FASTS (available at the European Bioinformatics Institute, http://www.ebi.ac.uk/Tools/sss/fastm/).  In addition to searching your organism database, consider searching a better annotated organism that is close evolutionarily.

ID mapping between different databases can be done with Uniprot ID mapping service:

http://www.uniprot.org/uploadlists/

Dear Hajer,

If you are using a model that is not well annotated, you can try searching the MS/MS RAW files against a database with the nearest taxonomy to your model if your model has no specific database. For instance, if you are working on guinea pig, you can set the taxonomy in the MS/MS search to "Rodents" or "mammalia". The search will take longer, and the file will be much larger and more complex, but you will be able to map the MS/MS peaks to the corresponding peptides and proteins in different species. It will not be easy, but since you are looking for certain substates = you have a target.