retrieving reads from SRA - lack of documentation

[NGSfan]

Is anyone else frustrated with being able to retrieve SRA submitted sequences?

Every time a paper comes out it seems like the SRA annotation is incomplete. There's no way to know which SRA ID # corresponds to which data set is mentioned in the paper (is it replicate 1? replicate 2? Illumina? SOLID? is it whole genome? is it whole exome enrichment?).

Is there problem with the submission form or process that makes it difficult? or are authors just being a little quick and careless?

I just don't understand why the process of retrieving sequence data has to be a sherlock holmes case every time. I have to write to the corresponding author, then track down the one author who is responsible, sometimes going back and forth between authors, then wait for he/she to sort it out.

This is not just one isolated case, but 5 times already.

My concern is that in two to three years there will be hundreds (or thousands) of publications with SRA entries, and it would be impossible to figure out which data set is which.

Are the microarray datasets also as difficult? Might we learn some lessons from microarray submissions?

[mgogol]

I haven't dealt much with SRA data, but microarray public data can be tricky too. Generally there is a file describing which files are which samples, but formats vary pretty widely. Sometimes you only get the raw data, and sometimes you get some "Processed" data, but the process isn't described in very much detail. I wish people would submit wig or bedgraph files to make it easy to take a look at the data without a lot of effort.

Having submitted microarray data myself, it's kind of complicated and I see why people are tempted to shove whatever works in there. I think this is a cultural thing that we need to change, but also could be changed top-down by SRA/Geo/ArrayExpress requirements.

GEO said if I make track files from a public project, I could re-submit them, and if the author approved, they could be added to the page for other people to download.

So maybe once you figure out which sample goes with which data set, you could send SRA another file to add to the project for those who come later.