1
$\begingroup$

I have a list of 100+ proteins and I need to conduct psi-blast for each one of them against "bacterial proteins" only.

I went to NCBI's protein blast tool, but couldn't figure out how to select/limit the targeted database.

Here are the 2 configs I tried:

  1. Database: pdb, Organism: bacteria (taxid:2)
  2. Database: landmark, Organism: bacteria (taxid:2)

Can anyone tell me if either of the approaches is correct? If not, please point the right way.


Edit 1 - add more details

All of my data is from "Supplementary Table 10" of paper "Comparative genomics of the neglected human malaria parasite Plasmodium vivax".

About 5% of all my protein sequences are ISS predicted. Remaining sequences are IEA predicted

$\endgroup$
  • 1
    $\begingroup$ pdb only contains IDs with deposited structures. I have not heard of landmark, could you provide a link to it? I agree with @Michael_A that Uniprot is the way to go. $\endgroup$ – James Oct 24 '16 at 5:13
2
$\begingroup$

Your approach is correct but it is worth considering which database will work best for you. Questions to consider are;

  1. Does the database provide a large search space for the organism(s) you're interested in?
  2. Do its IDs work well for searching other databases?
  3. How does the host database handle identifiers over time?

For protein sequences I would use;

Database; UniProtKB/Swiss-Prot(swissprot)

Organism bacteria (taxid:2) (or a subset)

The reasons for this are that I am comfortable with how Uniprot handles their ID's, they provide a large search space with includes clear delineation between reliable (reviewed) and other (unreviewed) sequences, and finally UniprotKB IDs map well to genes. The latter can be problematic. I also know that I won't have trouble searching PDB with UniprotKB IDs.

Despite using and liking the Protein data bank (PDB) I wouldn't use them for this kind of search as they are a structural database which reduces your search space. PDB has 8292 structures for E.coli compared to 23 017 reviewed and 1,335,860 unreviewed sequences in Uniprot. If you are only interested in structures then the PDB is an ideal fit.

I haven't used landmark.

$\endgroup$
  • $\begingroup$ I am looking for structural homology in bacterial proteins. Since you pointed out that PDB is a structural database, would that make it a better candidate? $\endgroup$ – zhirzh Oct 24 '16 at 5:16
  • $\begingroup$ I'd still use Uniprot as the uniprot IDs can be used to search PDB. However, you could do both. The only downside to Uniprot is that it will likely return a large list. Taking time to understand the structure of uniprot's holdings will help you to deal with that. If you can get away with reviewed sequences I'd stick with those. $\endgroup$ – Michael_A Oct 24 '16 at 5:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.