Sorting millions of snapshots from the Linac Coherent Light Source

August 31, 2011 By Glennda Chui
Sorting millions of snapshots from the LCLS
Randomly selected representatives from (a) nanorice snapshots; (b) Mimivirus snapshots; (c) miscellaneous snapshots; (d) blank snapshots; and (e) saturated snapshots, each from a separate bin of images sorted by the new, automatic method. Credit: Chun Hong Yoon, et al.

The great thing about SLAC’s Linac Coherent Light Source is that it churns out incredible volumes of data about things no one has ever seen before, such as snapshots of individual viruses.

The hard thing is: What to do with all that data? Of the several million snapshots scientists might get in a single, 10-hour shift of zapping samples with a powerful X-ray , fewer than 1 in 100 will contain the information they’re looking for.

This is the problem that Abbas Ourmazd and colleagues Peter Schwander and Chun Hong Yoon of the University of Wisconsin-Milwaukee are working on. They’re members of an 80-plus person collaboration, including SLAC scientists from the LCLS and the PULSE Institute for Ultrafast Energy Science, who performed groundbreaking experiments on large viruses and tiny protein nanocrystals with the world’s most powerful X-ray laser in December 2009.

In an Aug. 12 report in Optics Express, the team outlines a method for automatically sorting those millions of snapshots so most of the good shots end up in just one bin. Further analysis of the data in that bin should yield a 3-D image of the original object, Ourmazd said, whether it’s a virus or a tiny grain of “nanorice” that’s used as a test subject in these studies.

“Computer analysis – making sense of all this beautiful data coming out of the LCLS – has turned out to be as difficult and as important as the data acquisition,” Ourmazd said in an interview.

He describes the problem this way:

“Say, for example, you make a solution full of virus and you spit little droplets of the solution into the X-ray laser beam” with an injector that sprays them in a fine mist. ”Each droplet can be empty, or it can have one virus particle in it, or it can have multiple particles in it,” Ourmazd said. “Also, the solution is not necessarily pure, so you can have different kinds of particles being spat out individually or in combination.

“In addition, the X-ray intensity varies from shot to shot. Even when it’s the same, the laser beam may miss the particle, hit it square on, or just hit some of it and not the rest. What you would like to do is take these snapshots and from them extract the droplets that contain a single copy of a single virus which has been nicely hit.”

When the X-ray pulse hits the virus, it forms a diffraction pattern in a detector, and thousands of those patterns can be combined to get the structure of the virus. The challenge is to do this automatically, without direct human supervision; without bias; with high precision; and really fast.
Ourmazd, Schwander and Yoon are part of a theory group that is developing algorithms for analyzing the terabytes of data coming out of the LCLS. Their approach combines mathematics, scattering physics and information theory. Ourmazd is a distinguished professor of physics; Schwander, a senior scientist specializing in physics and informatics; and Yoon, a postdoctoral researcher and electrical engineer.

The simplest way to think about their method, Ourmazd said, is to start with the idea that experimental data is correlated – it hangs together, somehow – while background noise is random. You can think of the correlated data for each snapshot of an object, whether it’s a full-on hit or a glancing blow, as lying on a surface. “The interesting information content is in the shape and in the wrinkles of the surface,” he said. The question is, what is the characteristic wrinkle pattern and shape that distinguish one type of object from another, and good shots from mediocre ones?

The algorithm found these patterns of correlation and used them to sort 7,214 snapshots of nanorice and viruses. The results showed 90 percent agreement with sorting done manually by a human expert. The researchers estimate that a million snapshots can be sorted in less than 10 hours with this technique.

“You go to the data with no preconceived notions. All the information is in the wrinkles and the shape of the data set,” Ourmazd said. “It’s a wonderful combination of mathematics and information theory.”

Once you’ve done that, he said, you should be able to open the bin that contains full-on images of nanorice grains or viruses, taken from all possible angles as the tiny particles tumbled through the X-ray beam, and reconstruct their structure in 3-D.

Ourmazd said the team is already applying this approach to data from LCLS experiments, images taken with cryo-electron microscopy and free-electron laser studies of changes in individual molecules.

“We feel like kids in a toy store,” he said, “totally confused by the choices and overwhelmed with pleasure.”

Explore further: Lie, lay: a grammar trap lying in wait

Related Stories

Scientists make quantum breakthrough

April 20, 2011

( -- Scientists have demonstrated for the first time that atoms can be guided in a laser beam and possess the same properties as light guided in an optical communications fiber.

Recommended for you

Fusion reactors 'economically viable' say experts

October 2, 2015

Fusion reactors could become an economically viable means of generating electricity within a few decades, and policy makers should start planning to build them as a replacement for conventional nuclear power stations, according ...

Iron-gallium alloy shows promise as a power-generation device

September 29, 2015

An alloy first made nearly two decades ago by the U. S. Navy could provide an efficient new way to produce electricity. The material, dubbed Galfenol, consists of iron doped with the metal gallium. In new experiments, researchers ...

Invisibility cloak might enhance efficiency of solar cells

September 30, 2015

Success of the energy turnaround will depend decisively on the extended use of renewable energy sources. However, their efficiency partly is much smaller than that of conventional energy sources. The efficiency of commercially ...

Extending a battery's lifetime with heat

October 1, 2015

Don't go sticking your electronic devices in a toaster oven just yet, but for a longer-lasting battery, you might someday heat them up when not in use. Over time, the electrodes inside a rechargeable battery cell can grow ...

1 comment

Adjust slider to filter visible comments by rank

Display comments: newest first

not rated yet Aug 31, 2011
now that is beautiful CS project

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.