X-ray Crystallography is done in many labs around the world and supplying HCC labs, but not all, just those who're seeing this as fitting their research I suppose.

Considering that IIRC 1536 different filters are applied to shooting 1 image, it seems not difficult to create very large numbers. The image generation is highly automated.

Not heard a peep different of "end of Q2-13", but things have been going massively faster. Voodoo calc applied many months ago, we're now about 50 days away from steady supply [things started running faster now that "just" 7 sciences are active at WCG. The prediction was made at time of having 10 or 11 active... and participation seems to keep growing. Likely will more if the 7.0.xx client is endorsed and members discover the easy config ways, stead of that horrible app_info [app_config.xml can now be read in via the menu and acted on by client on the fly... personally tested the <max_concurrent> function... but as usual, wandering off topic]

Official word... the mod-admin will read and maybe polls the PI for a brief update.

i don't know what proteins are the subject of the study, but many are, at least partially, "intrinsically disordered" which means that they have no predefined topology or 3D structure. I'm not sure how that impacts the use of x-ray crystallography or if the technique just isn't used for proteins that have large, disordered domains.

Here is great article to read if you're interested in protein structure. It's only a few pages long and very comprehensible.

Don't forget you must divide the 461 million results (now 465 million) by the replication factor of two. So the number of images processed now stands at 465 million. Shooting multiple images of a single sample under different conditions, techniques or filters makes much sense from the scientific point of view.

Cheers

Good question, Hypernova. The replication factor and the number of work units per result have each changed over the course of the project, so the 461 million results do not translate directly to an image count.

The HCC images aren't actually diffraction images. They are images of a prior step in the X-ray crystallography pipeline -- crystallization trials. Whereas a small number of diffraction images are needed to turn a protein crystal into a solved structure, a typical protein crystal is the product of hundreds or thousands of crystallization experiments, each observed at multiple time points.

In the case of HCC, our images are from the High-Throughput Screening Lab at the Hauptman-Woodward Medical Research Institute at the University of Buffalo. Their main crystallization screen tests target proteins against a battery of 1,536 different chemical "cocktails". Each trial (one protein + one cocktail) is photographed between 6 and 10 times over a period of several weeks.

As NixChix suggested, these multiple conditions (sometimes including filters to control for lighting conditions) multiply the number of images generated per protein. I'm sorry I don't have an exact count right now. More images are being generated constantly. We'll put out an update soon that mentions the number remaining for this project.

twilyth, thanks for the article. Disorder in a proteins is a major difficulty in protein crystallization and structure solution. Often, specific disordered regions can be identified (or at least predicted) in the protein's amino acid sequence, and a modified version of the protein can be created with a better chance of crystallization and ultimately, structure solution. Sometimes (as in the article), co-crystallizing a protein with its interacting partner is the answer.


Christian A. Cumbaa
Research Associate, Ontario Cancer Institute

Thanks for all these clarifications. It is easy to see that the hundreds of millions of WU's are in the end perfectly related to the proteins investigated and does not seem anymore such an absurd number.

On the other hand it also shows that in this field of research to do a full screening when one considers so many factors and combinations of factors you go immediately to enormous numbers, and this justifies grid computing as the only acceptable way to run through such large numbers in an acceptable timeframe.
It also would justify GPU crunching as a way to manage larger projects 10 to 20 times the size of what just CPU crunching would allow.