2
$\begingroup$

It's easy to set the maximum E-value threshold in a BLAST request. I need to set the minimum E-value. Does anybody have any suggestions?

$\endgroup$
  • $\begingroup$ the only manner (of my knowledge) is to use a script (R, python, java) to do this $\endgroup$ – KingBoomie Dec 9 '16 at 20:35
1
$\begingroup$

There is no such option. You can perform the alignment and filter the hits later. However, I wonder why you really need that. Lower E-value indicates lower chance of an alignment randomly picked from a database, to have a given score.

$\endgroup$
  • $\begingroup$ I'm looking for records that are moderately similar, not highly similar, to my query. $\endgroup$ – Philip Heller Jul 13 '16 at 19:51
  • 1
    $\begingroup$ @PhilipHeller As I said, there is no option for that. You have to filter the results. If you have to look for similarity then you should look at percentage similarity/score instead of e-value. $\endgroup$ – WYSIWYG Jul 13 '16 at 20:30
  • $\begingroup$ I second WYSIWYG comment, I think it would be the easiest way. Extracting the results and filtering them using scripts could help you out a lot but would need a bit of time if you don't know any scripting langage. $\endgroup$ – Dart Feld Nov 10 '16 at 14:17
0
$\begingroup$

Setting a minimum E-value threshold will result in literally every single entry in the database with E-values above that value, as they will have very large E-values. You will expect a million hits with 1000000 E-value, and as a result you will not get any useful information from the search.

Can you explain why you want to set a minimum E-value?

$\endgroup$
  • $\begingroup$ I'm looking for records that are moderately similar to my query. Normally those records would be at the bottom of the hit list, but there are hundreds of thousands of records that are highly similar to my query and I'm not interested in them. $\endgroup$ – Philip Heller Jul 13 '16 at 19:50
  • 1
    $\begingroup$ If you want moderately similar hits can you not just use discontiguous BLAST or blastn with a custom matrix? $\endgroup$ – Joe Healey Jan 8 '17 at 23:28
0
$\begingroup$

A suggestion to get less hits is to take another database, for proteins you could use the PDB database and get orders of magnitude less hits.

Thinking a little bit more about your question: Blasting is probably not the right approach for whatever you want to do. Even if you manage to set a minimum e-value, you'll probably still get huge clusters of similar proteins around that value, and I guess that's what you're trying to avoid.

Blasting is really good to find similar proteins, if you want to get a cross-section of whatever-ases, you need to dig into bioinformatics a lot more.

$\endgroup$

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.