lifesciences.bg

Well, CD-Search is simply one of the great tools at NCBI. CDD itself is one of the great databases at NCBI. Especially the ability of the algorithm to detect domain features (like catalytic residues for example) is what makes it so valuable for me – you can produce valuable, sensible and confident annotation of your proteins with such a great ease.

CDD seems to be one of the projects that is actively being developed. It is now probably the best documented tool/database at NCBI. However, one important thing is missing in the documentation: how can you access this great resource programmatically?

Wait, wait… I know – you can start RPS-BLAST using the standard NCBI’s URL API:

$Bio::Tools::Run::RemoteBlast::HEADER{'DATABASE'} = 'cdsearch/cdd';
$Bio::Tools::Run::RemoteBlast::HEADER{'SERVICE'} = 'rpsblast';

but the output is definitely not what you may expect – this is a raw RPS-BLAST output, and NOT a CD-Search result with all those valuable annotations and domain feature predictions…

So how one can get a real CD-Search output? By experimenting with the POST parameters at the CD-Search web page I have found a simple way to do this.

All you need is to start a normal RPS-BLAST job using the URL API, e.g. using the BioPerl’s RemoteBlast class. However, do NOT get/parse the results – they are simply RPS-BLAST hits. Instead, save the Live Request Identifier (RID) of the job and do the following HTTP POST or GET request:

where  RID is the RID of your RPS-BLAST job. And you finally have the annotated CD-Search output in XML format, enjoy!

The magic CGI parameter here is “output=xml”. And as Larry Wall says, it is the magic that counts!