SRA metadata structure

[kovaliuk]

Hello,

We know from the SRA article in NAR that the SRA metadata model has 6 objects (http://www.oxfordjournals.org/nar/database/summary/1456)
i. Study – high level info e.g. goals of the study, literature references – may be linked to BioProject dbase
ii. Sample – may be linked to BioSample databases
iii. Experiment – library information; instrument information
iv. Run – library information; instrument information
v. Analysis – analysis results e.g. alignments, assemblies
vi. Submission – groups the other objects
And that: ‘A Study consists of a series of Experiments with unique Samples. The Experiment in turn, consists of a number of Runs.’
Our Big Question is:
What is the object of interest – what do people refer to – is it the experiment? The study? The run? We are interested in this because we are trying to map the data flows and so which object(s) should we be following?

We are also interested in:
useful diagrams or descriptions the way that the metadata in the SRA is structured, including the relationships between the metadata objects
Definitions of what a ‘run’ is.

[Torst]

It's good to see I am not the only one who gets confused with the SRA data model :-)

I tried to think of it as a tree, but it is really a forest/graph. We use the Study accession number in publications, but then we usually only have one Experiment.

How sure are you of the statement "A Study consists of a series of Experiments with unique Samples. The Experiment in turn, consists of a number of Runs." ?

I do know that a Run can contain multiple Samples. This occurs when you multiplex samples into a single lane on Illumina for example. The same Sample could also be in multiple Runs, say you sequence it with Illumina and 454. And so on.

[drdoodlebug]

The SRA submission portal is driving me over the edge. I have a massively multiplexed 454 16s pyrotag run, do I really have to create a biosample for every one of my 100+ samples??? Another question about naming objects, where does it ever end? I'm trying to put up a single run and I swear I've created a baker's dozen names for each step. There has to be a better system. I'm beginning to understand why I pull up American pubs that have their data deposited with EBI and DDBJ. Grrr!

[Torst]

The SRA submission portal is driving me over the edge. I have a massively multiplexed 454 16s pyrotag run, do I really have to create a biosample for every one of my 100+ samples???

Yes you do. But you can do it all quite easily using the "Generic.tsv" tab separated file you can create with a script or with some Excel wizardry. The online portal allows you to download it, or email sra@ncbi and they will send it to you.

For the SRA metadata, they have a 5 sheet Excel document you can fill out for bulk submissions too. Just email sra@ncbi. They have been very helpful for our bulk (>100 sample) submissions.

[drdoodlebug]

Thanks for the help. I'll give it a try.