Blast Xml For Multiple Databases

Question

[Yann]

Hello, I run blast2.2.25+ with multiple databases, I would like to view my result for each database in my XML result file, it is possible?

[Pierre Lindenbaum]

what do you mean with "view my result for each database" ?

[Lyco]

you could use table output instead of XML and do a grep.... Arrgh, no, pleeease don't kill me...

[Lyco]

you could use table output instead of XML and do a grep.... Arrgh, no, pleeease don't kill me... :-)

[Michael Schubert]

You could also use [?] on multiple xml files instead of one. Care to explain the reason why you want to have it that way?

[Lyco]

Yann should explain this him/herself, but I guess this question is about running a blast search against multiple databases in one go and then teasing out the individual databases later. I must admit that I don't really see a problem, as the database identifiers should be part of the hit entry name. In a table, this is easy to parse, and I am sure that the XML gurus can come up with some XSLT magic to this with XML, too. By the way, I see nothing bad in running blast with multiple databases - very efficient if there are multiple small databases.

Answer 1

[Nabellaleen]

If I remember, Blast can run a search on only 1 database at a time, no ?

To run a Blast on multiple databases, you have to merge them with blastdb_aliastool, which create a "new" database. That's "forbid"/"block" the possibility to distingue initial database used for each result.

If you give multiple databases directly to Blast, I don't know what happened, this is not documented (I didn't found an answer anyway). So I think if it doesn't crash, it probably call blastdb_aliastool.

In any case, on NCBI, Blast XML output is "based" on this DTD : http://www.ncbi.nlm.nih.gov/data_specs/dtd/NCBI_BlastOutput.mod.dtd

It says we have a "BlastOutput" element (<!ELEMENT BlastOutput ( ... )>) contains a "BLAST Database name" attribute (<!ELEMENT BlastOutput_db (#PCDATA)>) and is a parent of iteration elements (<!ELEMENT Iteration_hits (Hit*)> - one by sequence in the database), which is a parent of hits (<!ELEMENT Hit ( ... )>) which are parents of HSPs (<!ELEMENT Hsp ( ... )>)

So, if you have multiple databases in an XML result file, there is a "BlastOutput" element for each database used, which contain the DB name in "[?]Your DB Name[?]" element, and which also contain results on this database in "[?]" element.

[Lyco]

@Nabella: what BLAST are you referring to? I routinely use blast with multiple databases (although I tend to avoid XML output) the syntax is -db "data1 data2" for new BLAST+ or -d "data1 data2" for traditional blastall.

[cmdcolin]

It is definitely possible to do this without aliastool, as mentioned above, passing multiple arguments for db. It has the annoying consequence of not telling you "which" database a particular sequence came from though

[Pierre Lindenbaum]

@Nabella, As far as I can see in Yann's previous questions ( Yann ), he already knows how is structured XML blast. So, I still don't understand what he means with "view"

[Nabellaleen]

My comments are based on my (failing ?) memory for Blast command line use, so ... If you say that's possible, I believe you. But my others comments stay : how blast treat this multiple databases ? What are the results ? ...

However, as Pierre said, maybe that's another subject ;)