Comments on: Do we need an X-ray Diffraction Image Data Bank?

By: Jan Dohnalek

Jan Dohnalek — Mon, 08 Jun 2009 06:12:09 +0000

Definitely YES. We have been discussing this in the computational work package of Instruct as well (Integrated infrastructure for structural biology in Europe). Amounts of data, transfer times and standardization are the real bottlenecks. There “might” be personnel dedicated under INSTRUCT in the future.

Jan

By: Ezra Peisach

Ezra Peisach — Sun, 07 Jun 2009 15:22:12 +0000

Such a resources would be useful for archival purposes as well as an aid for software developers.

There are a few major hurdles – including the meta information.
From a point of view of indexing processing – all the various criteria would be necessary to accurately reproduce the experiment. This would include spots used for indexing, I/sigma cutoffs, resolution limits in various programs, frames used for indexing, etc… The list goes on. For the well diffracting crystals – all of the above is not necessary – but for those marginal crystals – having the above information will be necessary to “reproduce the experiment”

Unless there is buy in from equipment manufacturers to dump this information from the data collection/processing – it will not happen.

Then there is the issue of cost. At the Tardis site – they report 33 datasets for 16 structures coming out to 106Gb (3.2Gb/dataset). From the PDB website, in the first quarter of 2009, 2041 structures were deposited. So say 8000 per year… This is about 50 terabytes per year of data!!! This is without mirroring, etc. Perhaps there is a solution with regional databases – by country, region, etc. – but this would require government support…

So – I see the value, there is a hefty cost…

By: DrNO

DrNO — Sat, 06 Jun 2009 17:55:05 +0000

What about space and upload/download time? A data set can be a several GBs. With 5-10k? structures solved/year, that’s 500 TB. I’ld say let’s wait a few years. Let us first have a unified system for images, something that would convert the detector specific image to a general image format that would be easily recognized by all indexing programs. I hate having to ask HKL2000 every time I use a new detector.

By: Sean

Sean — Sat, 06 Jun 2009 01:07:57 +0000

James, awesome.

I will be looking forward to the tools being released. I presume that the reason the depository has been used generally by the Australian community is due to being developed there?

I did not come across a reason why this could not be expanded into a world wide depository of diffraction images.

By: James Whisstock

James Whisstock — Fri, 05 Jun 2009 23:11:15 +0000

Check out http://www.tardis.edu.au/ which provides a practical distributed mechanism for deposition of data – currently this is generally used by the Australian community, however, the download tools that will be released shortly are designed both to permit ready archiving of data (a useful thing in its own right) as well as linking these data to the repository.

Cheers

By: David Waterman

David Waterman — Fri, 05 Jun 2009 17:49:11 +0000

Yes, it would be very useful, for developers interested in integration methodology and detector properties as well as the reasons mentioned above.

By: Graeme Winter

Graeme Winter — Fri, 05 Jun 2009 15:41:24 +0000

This keeps on coming up, I keep on saying “yes” we need to be able to store / share diffraction data. Some criteria should be applied on the amount of metadata available, for example the intention of the data collected, the numebr of sites / sequence / MR model so that the structure solution may be repeated.

By: Sean

Sean — Fri, 05 Jun 2009 14:38:26 +0000

I agree that an X-ray data bank won’t solve the problem with synthesized data.

My thought is that very few will even submit synthesized images. The reason being is that eventually they will be caught. If the structure has been published then at some point it will be attempted by another researcher.

This is an interesting point, but one that should not prevent a data bank from being developed if the community would find it useful.

By: Oliver Clarke

Oliver Clarke — Fri, 05 Jun 2009 06:18:40 +0000

Unfortunately, I doubt it’s going to solve the unfortunate issue of synthesised data – if you’ve seen any of James Holton’s simulated diffraction images, they look astonishingly realistic (both visually and with regards to the intensity stats, errors etc).

Additionally, that particular structure is unquestionably a fake – one can see that just by looking at the chemically impossible contacts that are apparently perfectly well-defined in the electron density.

The fact that Nature published the (blatantly inadequate) reply by the authors and didn’t request retraction of the paper was a little disappointing, to say the least.

(And if you have a browse through some of the previous structures that Krishna Murthy has published, it would appear not to be an isolated incident…)

By: Oliver Clarke

Oliver Clarke — Fri, 05 Jun 2009 06:10:23 +0000

Absolutely we need this as soon as practical, for all the reasons Sean outlined above.

Additionally, to aid interpretation of unmodelled density, all known and suspected components of the crystallisation solution should be listed with the deposition (not just the precipitant solution).

Preferably one would also include a direct beam shot or the coordinates of the refined beam centre.