Skip to Content

You Call That An X-Ray Source?

Over the years, when some puzzling feature of a drug candidate’s binding to a target came up, I’ve often said “Well, we’re not going to know what’s happening until some lunatic builds a femtosecond X-ray laser”. Various lunatics are now pitching in to build some. I’m going to have to revise my lines.
The reason I’d say such a mouthful is that we already, of course, get a lot of structural information from X-ray beams. Shining them through crystals of various substances can, after a good deal of number-crunching in the background, give you a three-dimensional picture of how the unit molecules have packed together. Proteins can be crystallized, too, although it can be something of a black art, and they can be either crystallized with or soaked with our small molecules, giving us a picture of how they’re actually binding.
There are, as mentioned earlier around here, plenty of ways for this process to go wrong. For starters, a lot of things – many of them especially interesting – just don’t crystallize. And the crystals themselves may or may not be showing you a structure that’s relevant to the question you’re trying to answer – that’s particularly true in the case of those ligand-bound protein structures. And the whole process is only good for static pictures of things that aren’t moving around. It used to take many days to collect enough data for a good crystal structure. That moved down to hours as X-ray sources got brighter and detectors got better, and now X-ray synchrotrons will blast away at your crystals and give you enough reflections inside of twenty minutes. And that’s great, but molecules move around a trillion times faster than that, so we’re necessarily seeing an average of where they hang out the most.
Enter the femtosecond X-ray laser. A laser will put out the cleanest X-ray beam that anyone’s ever seen, a completely coherent one at an exact (and short) wavelength which should give wonderful reflection data. The only ways we know how to do that are on large scale, too, so it’s going to be a relatively bright source as well. The data should come so quickly, in fact, that several things which are now impossible are within reach: X-ray structures of single molecules, for one. X-rays of things that aren’t in a crystalline state at all, for another. And femtosecond-scale sequential X-ray structures – in effect, well-resolved high-speed movies of molecular motions.
Now that will be something to see. Getting all that to work is going to be quite a job, not least because X-ray bursts of this sort will probably destroy the sample that they’re analyzing. But there are two free-electron X-ray lasers under construction – one set to complete next year at Stanford’s SLAC facility and a larger one that will be built in Hamburg. “Large” is the word here. The smaller SLAC instrument is already two kilometerslong. According to an article in Nature, though, a Japanese group have proposed some ways to make future instruments smaller and more efficient – all the way down, to, um, the size of a couple of football fields. But there’s another completely different technology coming along (laser-plasma wakefield instruments) that could produce far shorter X-rays in one hundredth the space, which is more like it.
I don’t think we’re going to see a benchtop-sized X-ray laser any time soon, especially since these things are going to need to be large just to get up to the brightness that will be needed. But I’m very interested to see what even the first generation machine at Stanford will be able to do. There are a lot of mysteries in the way that molecules move and interact, and we may finally be about to get a look at some of them.

12 comments on “You Call That An X-Ray Source?”

  1. d_orbital says:

    Wouldn’t that be a Xaser? 🙂

  2. Sili says:

    Sadly I’ve only ever heard it referred to as FEL (like the swoop).
    I guess I should get off my lazy (depressed) arse and learn some black art while there’re still jobs to be had.

  3. Hap says:

    I thought X-ray/free electron lasers were originally intended for Star Wars – lab-scale FEL might not be an unmixed blessing.

  4. Jose says:

    IMMA FIRIN MAH LAZER!!
    (apologies, couldn’t resist).

  5. Larry Evil, PhD says:

    Dear NIH,
    Please find attached my grant proposal for the funds necessary to purchase 10 (ten) frickin sharks.
    Sincerely,
    Larry Evil, PhD

  6. alchemist says:

    I wonder how much one run of a sample will cost. Where is all this energy coming from to run this thing?

  7. SRC says:

    Just out of curiosity, what do they plan on using for a beam stop?

  8. MolecularGeek says:

    Just out of curiosity, what do they plan on using for a beam stop?
    Venus?

  9. Oli says:

    I don’t think they use a beamstop, as a rule (or at least not a conventional one before the detector). Mostly you either use a special detector with a central hole, or use a mirror with a central hole to direct the diffracted x-rays to a conventional detector off the beam axis.

  10. Brooks Moses says:

    Well, SLAC is aimed into the side of the Santa Cruz mountains, pretty much. Or out of them; I’ve never actually known which way the beams accelerate there. I suppose that would probably work for a beamstop.

  11. David Pearlman says:

    Boy, every few years we hear this kind of news again. It goes back to the early ’80s, when they reported on a tuned laser that was going to eliminate the phase problem in x-ray crystallography. Or not.
    At any rate, here is the hidden problem here: Assuming we take everything else at face value, and accept that they can obtain data that reflects a single femtosecond snapshot, the process of going from the experimental data to a “structure” is still going to be one that is underdetermined and in part ambiguous. The solution will be, as it is for crystallography now, to use computational chemistry to complement the experimental data. From the mid ’80s onward (when Xplore first became popular), nearly every x-ray crystallographic structure has, in fact, been a structure suggested by x-ray data and smoothed out by the energy function in the (minimization/MD) refinement program. The result is something that looks kind of like the “real” structure (whatever that means when you’re talking about a single structure for data collected in a non-instantaneous manner), but which also reflects, sometimes to a non-trivial degree, the energy function in the refinement program.
    Bottom line is: Even if we could reduce the timescale of the measurement process from the current hours or days to 1fs, it’s not clear how much better the “structures” would be, given the reliance and biases introduced by the energy function.

  12. befuddled says:

    For more information on this subject, I recommend Ed Lattman’s overview at (gasp!) PNAS:
    http://www.pnas.org/content/98/12/6535.full
    I don’t see how you can get “femtosecond-scale sequential X-ray structures” though. The first pulse will fry the molecule under investigation. But gathering a *large* number of structures should delineate the most populated conformers.

Comments are closed.