For doing data analysis, it’s much easier than booking time on a cluster at work, as I can install exactly what I want without having to go through the cluster admin and I don’t have to wait in line. Unless we start doing a lot more data analysis, I think it’s still going to work out much cheaper than buying in our own hardware and it’s certainly going to be less hassle. I’m envisaging just running instances of the Galaxy and analysis AMIs when we’re actually doing the alignment, peak finding and so on. I’ll probably keep the short read data as BAM or BioHDF on S3 so people can get at them and run their own analysis on EC2 if they want to.

I also need to set up a LIMS-type server for managing and sharing our data. Probably with some kinda REST-API so other stuff (network building tools etc) can query the data (eg. “What experiments have you done that involve NRSF? as RDF?” or “Give me all the features between these chr co-ords for NRSF binding in NS5 cell lines as BED, or SAM or something). We can host this kind of thing at work and it doesn’t really need to scale much. So I imagine I’ll have the analysis pipeline hand me back a list of peaks (genome pos, plus score, pvalue etc) and some metadata describing the analysis workflow and dump that into my LIMS thing.

I’ll bring my laptop to the biogeeks thing if you want to have a play with the AWS stuff.

I guess I’ll probably just stick the short read data onto EC2 for analysis and have it spit out the binding peaks and analysis metadata back to me.