Data retention policy for Solexa

[bioinfosm]

With the improving GA and more data coming off the instrument, it becomes even more important to be efficient in what to archive. I think there can be a core set of intensity values and matrices, which can be used at any time with the same version of software to generate the same data. This will obviate the need to save the entire analysis folder.

Obviously, the assumption here is that images are too large and need to go as soon as they are converted to intensities.

Any thoughts on this core set of data to back up, which can be later on expanded to the entire run folder for any different analysis? ALso, any other ideas of archiving the 'useful' bit of data off of a solexa run?

[new300]

It should be possible to reconstruct everything from the raw intensity and noise files in the firecrest directory. If you're recalibrating you'd also need to raw data from the control lane as well. So if you store all the raw signals from a run you should be able to reconstruct the processed in intensities, 4 quality scores and basecalls, you'd need to hack the pipeline to do this (or use Swift or another basecaller) but it's entirely possible.

The crosstalk matrices are calculated from the data itself, so don't need to be stored (though they are small anyway) as are the phasing corrections. The offset matrices are only useful during image analysis. You could store a subset of data from the runlogs which contain Z height and other run information, however this information isn't currently used in my the primary data analysis so isn't really critical.