IUCr forums

(Posted on behalf of Tom Terwilliger for technical reasons.)

This is a summary of some of the unique points made in a discussion on archiving images for PDB depositions Oct-Dec, 2012 on the CCP4 bulletin board. The IUCr has commissioned The Diffraction Data Deposition Working Group (DDDWG) to consider these issues. There is a public DDDWG web site where issues are discussed.

T. Terwilliger

Relevant topics on the CCP4 bulletin board include:

• IUCr committees, depositing images
• IUCr discussion forum on diffraction data deposition
• raw data deposition
• raw data deposition (off-list)
• To archive or not to archive, that's the question!
• Archiving for fraud detection
• Archiving Images for PDB Depositions
• Image compression

Unique points made during the discussion

Information contained in raw images but not processed data

• Diffuse solvent contributions, commensurate and incommensurate superstructures, split reflections
• Multiple lattices
• Resolution of the data

Usefulness of raw images in reinterpretation of the data

• The data can be reanalyzed and reinterpreted
• Data can be reprocessed with greater care or skill than was applied the first time
• New biological insight can be obtained by reanalysis
• There are many examples where re-interpretation changes conclusions (somewhat)
• A completely different biological conclusion can be obtained by re-refinement
• Saved images allow systematic re-analysis of all deposited data

Improvements expected by saving raw images

• Better data processing software will certainly be developed and will provide a better data model, allowing better structure models.
• Improved methods can be applied for removing ice rings from images
• The raw images are a permanent experimental record
• Raw images make a rigorous check by reviewers possible
• Raw images are highly useful for methods developers
• Raw images allow re-analysis of modulated structures
• Saving raw images is the next logical step after saving structure factors
• Solving structures by completely new methods may require the raw data
• Finding over- or under-merged datasets requires the raw data

Disadvantages of saving raw images

• We may move totally beyond today's crystallography in the future and all this will be moot
• The effort would be better spent to save the crystals instead
• Perhaps we should focus instead on storing DNA
• Maybe the question is not "To archive or not to archive" but "What to archive"
• A view of one who is unconvinced of the utility
• And a rebuttal...
• It takes a lot of effort to store data in an organized way
• Preparing data for standardized deposition is time-consuming

Feasibility and costs of raw data storage

• Estimate of cost of data storage
• The cost of data storage is decreasing a factor of two every 14 months
• Storage is already being done at synchrotrons is
• Only saving a few datasets will be necessary for each structure so the amount of storage will not be huge.
• Local storage of images is entirely feasible and should be done routinely
• Who will pay for the costs?
• It is important to compare the costs of storage of data with the costs of repeating the experiment

What data should be saved?

• Save raw images through processed data along with metadata in a systematic way
• A systematic scale for data to be saved
• Not-quite-solved datasets would be useful to save in addition to solved datasets
• A focus on the datasets representing deposited structures improves practicality

Document identifiers and standardization

• A practical approach to storage of images is assigning DOI's to data and storing locally (e.g. at synchrotrons)
• DOI's can be assigned locally and saved as part of deposition; this can be tested systematically
• It is important to have standardization on image headers for re-useability
• Making sure data is safe
• Alternative storage locations may be useful

Examples of archiving images and metadata

• The IUCr is collecting examples of data that could be deposited and examining whether archiving as part of publication could be carried out
• TARDIS example of federated data storage of raw images
• The JCSG database is an example of how all this could be done

Practical issues

• A detailed analysis of the practical issues in storing images
• As a first step, we should save unmerged data
• Should release of raw data be compulsory?
• Who will ensure that the images deposited correspond to the structure deposited?
• A repository of data makes it possible to find images when you need them
• Saving images for fraud detection requires knowing the last trusted computer

Image compression

• Perhaps we should save compressed images
• Would people use compressed images?
• No need for compression

Crystallographers' opinions

• An informal poll on saving diffraction images