Minimal RNA seq dataset for performance eval

[obscurite]

Hello all. I'd like to compare the speed for the tophat/cuffdiff protocol on various hardware platforms to see which is the speediest combination. Is there a small genome/sample set that anyone knows of that would be ideal for testing? The run should be quick (no more than 15 minutes per run) and ideally produce some differential expression data or at least be verified.

My default approach is going to be to use the genome I already have for my samples and then just chop off a small piece of left and right pair end reads (probably 10MB worth or so). I doubt I'll see anything interesting in the output but at least it will be a speed test.

Ideas are welcomed. Sorry if this question is naive, I'm just starting with RNA seq. Any information on speeding on rna seq would be welccomed. I just found STAR which looks interesting, but don't know much about it (http://gingeraslab.cshl.edu/STAR/).

[GenoMax]

You could use the dataset suggested by Cole in their protocol paper: http://www.nature.com/nprot/journal/....2012.016.html. If you down sample it then you may or may not get interesting results from the set.

STAR is fast and is the choice of aligner several people use. If your genome is large then be ready to have ample memory available for STAR use.

[obscurite]

Thanks GenoMax. It so happens that's the paper and protocol we started with, so I've considered using the iGenome fly data, but the problem is the genome is still hefty as are the data files. I think they compare with a normal RNA seq data set. The yeast genome is smaller, but I couldn't find the published diff. gene data like they had for the fly.

I was just in touch with the author of STAR and will be trying it this week. We have two platforms, one has 20GB/4cores and one has 24GB/8cores. If we can run tophat, I'm assured STAR will be much faster. It should be educational to compare the data that's generated.