[ScottC]

They're storing the images?! Why? That's a serious amount space to assign just for archive. AFAIK this wasn't done routinely for ABI sequencing, so why do it for HTS? Is it a justifiable expense in case someone would wish to re-analyse them?
al...

I was considering what our data retention policies will be when we get our system fully up and running to capacity. Naturally, I tended to think that deleting the images was best just because they occupy so much space. Once you have the base calls, they're not much use... but then I heard a few talks regarding improvements to the Illumina software that does the channel deconvolution and how these improvements might lead to better base calling etc. Well if the images are gone, you'll never have the chance to get that improved data. But then, will anyone even want them reanalyzed. I guess it's a trade-off between storage cost and estimated future value. Maybe it's just easier to do the run again rather than reanalyse old images.

At a recent workshop I attended this was a common query and according to some accounts current de novo software can't cope with the depth of coverage generated by Solexa et al...

I guess that's true in some respects... but it depends on the software you use. I know a lot of 'old' software can't handle it, but there are new ones now that can (Velvet and all the rest). We're doing a bit of work on that ourselves. I think the problem is more the short reads than the depth of coverage. Well, getting back to my original gripe... I think the reason that there's so much human- and mammal-centric work going on (rather than my favourite - bacterial stuff) is that there's more money in it :-)

[chris]

Hi Scott,

Once you have the base calls, they're not much use... but then I heard a few talks regarding improvements to the Illumina software that does the channel deconvolution and how these improvements might lead to better base calling etc. Well if the images are gone, you'll never have the chance to get that improved data. But then, will anyone even want them reanalyzed. I guess it's a trade-off between storage cost and estimated future value. Maybe it's just easier to do the run again rather than reanalyse old images.

That's my point. If it's going to n thousand $currency to store the images on the off-chance that someone may want to re-analyse the base calls it may just be cheaper to re-run the experiment - assuming you still have samples of course

What kind of improvements in the base-calls are we talking about and how much of a difference will it make to a final assembly?

Well, getting back to my original gripe... I think the reason that there's so much human- and mammal-centric work going on (rather than my favourite - bacterial stuff) is that there's more money in it :-)

Well, that's always going to be the case isn't it. However, smaller genomes will benefit most from this type of data as there's a much greater chance of unique reads. Lower costs also mean better chance of getting funding for the sequencing.

[JohnBull]

eh?

Something not often seen in molbio research space. The process supervisor really put together a nice story about their pipeline and how they handle and track seemingly "routine" processes on a scale that is like no other in the world (20+ Solexa GAII-PE machines!).

GAs

Sanger=28
BGI=19

!

[ECO]

Ok ok..._almost_ like no other in the world.

Apologies to any sensitive Sangerites out there.

We'd love to hear about the LIMS and data management pipeline in use there too!

[ScottC]

That's my point. If it's going to n thousand $currency to store the images on the off-chance that someone may want to re-analyse the base calls it may just be cheaper to re-run the experiment - assuming you still have samples of course

What kind of improvements in the base-calls are we talking about and how much of a difference will it make to a final assembly?

I'm not sure at this point, but I do know that there are a few packages on the horizon that will produce new base calling results. I guess we'll have to wait and see as to how good they are, and whether it's worth keeping all that data.

Cheers,
Scott.

[ScottC]

We'd love to hear about the LIMS and data management pipeline in use there too!

Yeah, definitely! Post post! :-)

[ScottC]

That's my point. If it's going to n thousand $currency to store the images on the off-chance that someone may want to re-analyse the base calls it may just be cheaper to re-run the experiment - assuming you still have samples of course

What kind of improvements in the base-calls are we talking about and how much of a difference will it make to a final assembly?

I'm not sure at this point, but I do know that there are a few packages on the horizon that will produce new base calling results. I guess we'll have to wait and see as to how good they are, and whether it's worth keeping all that data.

Cheers,
Scott.