SRA format

[Deutsche]

Does anybody know about the SRA format specification? Does one exist?
I just have found API on the NCBI site which can help to read SRA files but I haven't found any information about the format specification.

[vadim]

ask them: [email protected]

[maubp]

You don't mean the SRA XML Specification, which is documented?

http://www.ncbi.nlm.nih.gov/books/NBK49285/

Rather I assume you mean the binary SRA files whose first 8 bytes are "NCBI.sra"? If you find a link, or you prompt the NCBI to publish this, could you post the URL here please?

[Deutsche]

You don't mean the SRA XML Specification, which is documented?

http://www.ncbi.nlm.nih.gov/books/NBK49285/

Rather I assume you mean the binary SRA files whose first 8 bytes are "NCBI.sra"? If you find a link, or you prompt the NCBI to publish this, could you post the URL here please?

Yes, I mean the binary file format. Ok, I will write them and will answer you if I find something.

[vadim]

You could also check their source code of, say, fastq-dump.c or of some other dumping tool. It worked for us.

[Deutsche]

You could also check their source code of, say, fastq-dump.c or of some other dumping tool. It worked for us.

It is very difficult to understand a format specification from 12 Mbs of code so I want to try the simplest way at the beginning. If nothing is successful I will try to analyse the source code.

[Deutsche]

ask them: [email protected]

Could you tell me please how did you got this address?

[vadim]

Could you please explain what are you planning on doing with SRA data? Most people are happy with fastq/fasta dumps produced by standard tools. For something more complicated you could use the API from the SDK, which in my sense is much easier than understanding the format specs.

[Deutsche]

Could you please explain what are you planning on doing with SRA data? Most people are happy with fastq/fasta dumps produced by standard tools. For something more complicated you could use the API from the SDK, which in my sense is much easier than understanding the format specs.

I'm working in the UGENE project and my next task is integration SRA format supporting into our tool. It is not simple to have just included SRA SDK into UGENE because of our tool is a cross-platform program but this SDK is only UNIX-supportable.

[vadim]

I believe SRA SDK can be build for Windows and Mac as well, although I have never actually tried this.
Is UGENE written in C++? In which case I would definitely consider re-using NCBI's code.

[Deutsche]

I believe SRA SDK can be build for Windows and Mac as well, although I have never actually tried this.
Is UGENE written in C++? In which case I would definitely consider re-using NCBI's code.

Yes, it is. C++ with Qt4.

[Deutsche]

If you find a link, or you prompt the NCBI to publish this, could you post the URL here please?

Guys from NCBI said me that they don't give this documentation anybody. And if you want to use the SRA format then you need to use their API.

[maubp]

Guys from NCBI said me that they don't give this documentation anybody. And if you want to use the SRA format then you need to use their API.

Well, at least they are clear about it.

Hurrah for the principles of openness and sharing in science! </sarcasm>

[jkbonfield]

It's perhaps a reasonable stance to take as it gives them flexibility of changing the format without having to keep notifying people, just as long as the API remains constant. However it does rather block interfaces being written by others in alternative languages.

The format is almost certainly quite complex though. I remember lots of discussions and to-ing and fro-ing on the best algorithms for compressing traces, qualities and sequences, with different methods for each type. As others have suggested, I'd recommend using their API and if it doesn't port cleanly to Windows then making it fixing that may be an easier task than reimplementing.

I'm not sure what licence they use though and whether that would be a hindrance.

[vadim]

As others have suggested, I'd recommend using their API and if it doesn't port cleanly to Windows then making it fixing that may be an easier task than reimplementing.

I'm not sure what licence they use though and whether that would be a hindrance.

It should work under windows, see here:

http://trace.ncbi.nlm.nih.gov/Traces/sra/sra.cgi?view=software

It is not licensed, I asked them recently about it and they said "no restrictions", whatever that means.