Skip to Content

The X-Ray Future, And Others

It’s been up for a while, but I found this essay on the future of X-ray crystallography at Chemical and Engineering News to be interesting. It was part of their “100 years of crystallography” section from August 2014, and those sort of have to end with a “Where do we go from here?” piece. Even if you’re not into X-ray, though, it’s worth a read, because the issues it raises are experienced by many other fields.

That is to say, they get easier, and as they do, they work on harder problems. Back when the first inorganic salts were being obtained, getting any x-ray structure at all was an invitation to lots of hard work. Just gathering the reflection data was (by modern standards) a terrible slog, and then you had the fun of doing all the calculations by hand. It gives a person the shakes just thinking about it. Under those conditions, working out the crystal structure of copper sulfate was a real accomplishment. (And just think of the nerve of pioneers like Bernal and Hodgkin, going after protein crystals at such early dates).

It got better. And it got better from many directions at once – better computational methods, wildly better computational hardware to run them, far brighter and higher-quality X-ray sources than anyone could have imagined in the old days, and hugely more sensitive detectors. All those processes are still going on, by the way. What that’s meant, as it has in any other field (think DNA sequencing for one example) is that what was first worth a Nobel then became worth a PhD dissertation, then became just some of that grunt work that had to be done on the way to something else, and then eventually became a routine method that involved turning a couple of knobs and hitting a few keys. All kinds of small-molecule X-ray structures are routinely obtained by people who can barely begin to explain any of the math or physics behind them, in the same way that you wouldn’t want to sit a thousand walk-up NMR users down, give them blank pieces of paper, and ask them to consider the Hamiltonian (shiver).

The people who actually lived through the heroic era of any technology can react to all this in several ways. Some get bitter – there were, I’m reliably told, organic chemists who had a terrible time getting used to what NMR spectra could reveal about natural product structures, because they were so heavily invested in the old methods of structure determination. Others follow Elvis Costello’s advice, and thus used to be disgusted, but now try to be amused. And others just look at the way things used to be, are deeply happy that neither they nor anyone else has to go through that any more, and turn their attention to the sorts of problems that are on the current edge of feasibility. There’s always hard stuff out there if you want it.

For crystallography, some of that hard stuff is to be found with bigger molecules, smaller samples, and shorter collection times, and the end result of that would be atomic-level resolution of huge protein complexes, in situ, on a femtosecond time scale. Can that be done? Got me – I’m glad to don’t have to deliver it. But if you push all the sliders over all the way, that’s what you come up with, and no one can say that it’s completely impossible, either.The sorts of things being done now would have been thought pretty unlikely not all that long ago. Doesn’t have to be done with just X-rays – you can diffract electrons, neutrons, what have you, and maybe you’ll want all of the above in the eventual wonder machine.

What that means, though, is the same as in any other field as it evolves: it’ll move away from you if you’re not careful. You have to keep a close eye on things as what used to be special gradually becomes routine, and make sure that you’re not just offering your services to the diminished number of people who haven’t caught on yet. That, actually, is what I was getting at in those posts about the “synthesis machine“. As they exist now, no one’s automated organic synthesis device is much of a threat to the field’s practitioners. But it doesn’t have to stay that way, and a lot of people are trying to make sure that it doesn’t (and I think that they have the right idea). In the sciences (and in most other situations, too) work that can be done by machines probably should be done by machines, and free the rest of us up to do something else.


20 comments on “The X-Ray Future, And Others”

  1. Kelvin Stott says:

    I did my PhD on protein crystallography in the 1990’s and worked directly with Max Perutz on a joint project. He told me plenty of great stories, including the fact that he had an army of attractive young women feeding in punch cards to solve the structure of hemoglobin.

    I also fondly remember that he would peer through the little window into the cold room where I spent hours setting up (litterally) thousands of crystallography screens. He would have to stand on his tip toes because he was quite short, and would steam up the window with his warm breath. Those were the days, I still miss them (and Max himself).

  2. anon says:

    We shouldn’t forget Linus Pauling and his contributions to crystallography. In addition to his work with proteins and inorganic salts, he studied structure determination of molecules in gaseous and liquid states.

    1. anchor says:

      Pauling..Yes and also not forget the father and son Bragg from UK and if am right they both won Nobel!

    2. cancer_man says:

      My college physics lab was x-ray crystallography. I remember reading an interview when Pauling was around 90 in Scientific American (I think, but that could be wrong) where he said he preferred talking to undergrads since they had more interesting ideas and that he thought researching x-ray crystallography was boring.

      Has anyone read that as well? curious.

  3. steve says:

    And soon, with advances in cryo-EM, X-ray crystallography will be remembered the way records are now remembered before the advent of MP3 players. And so it goes….

    1. AGMMGA says:

      For big protein complexes, maybe, and as a crystallographer I say about time. But for smaller stuff, I doubt it. Crystallizing a single protein / domain + soaking 200 inhibitors + solving their structure is very fast. Crystallization is a limiting step, but once you have that down pat, the rest is peanuts.
      Cryo-EM reconstructions are insanely slow, both in terms of data collection and data processing. Even allowing for increase in detector sensitivity that might routinely push size down to 1 month of data collection + number crunching per structure… not practical at all.

      1. AGMMGA says:

        * 1 month of data -> 100 kiloDalton (and that would be a big step), you are still looking at 1 month of data collection

        Somehow a line was lost in my reply…

        1. steve says:

          I’m certainly no expert in the area but what you describe seems like it should be amenable to technological improvements, e.g., better number-crunching algorithms, more computer power, etc.

        2. VinceC says:

          Data collection for cryo-EM is not the problem. You can collect sufficient data in 24-48 hours on a high resolution instrument (300 kV source and direct electron detector) to solve the structure. The bottlenecks (from my limited perspective) are sample prep (no high throughput methods as in crystallography), data analysis, data storage (massive data), and methods for judging goodness of fit of the model to the density map. That being said, this field is undergoing a revolution.

  4. Magrinho says:

    Happy Friday! Elvis Costello and, inadvertantly, Nick Lowe (And so it goes) all in one morning.

  5. Hap says:

    I think I’d be happier with the consequence of mechanizing easier work and of getting freed up to do more valuable and interesting work if that were the case. Most of the people who have done the easier work, though, probably won’t be doing the hard and interesting work – they won’t be doing anything (at least not near what they did before). In some cases, the hard and interesting work doesn’t get done at all, because it’s not worth it. And since we are not good as a society of giving people the ability to use their knowledge to do something else, most of those people will not end up doing anything useful with that knowledge. That’s why people fear the evolution of the workplace and unemployment, because their ability to do what they know how to do will not change, but disappear.

  6. Ash (Curious Wavefunction) says:

    The real deal might be Feynman’s prediction: simply looking at structures atom-by-atom through scanning tunneling or force type microscopy techniques (which as detailed on this blog have already made a dent on simple structures). If such techniques become general, robust and high-throughput enough that would be the end of conventional crystallography. An even more audacious application proposed by Feynman was synthesis by scanning force microscopy. Perhaps the physicists would have the last laugh after all.

  7. JH says:

    Difficult to see how AMF would ever deal with structures that are far from planar. No matter how you orient the molecule, there are going to be atoms inside as well…

    1. Mark Thorson says:

      Simple. You image the atoms on the outside, then you rip them off and scan what’s under them. It’s like taking apart a car engine. First you take off the air filter, then the carborator, then the intake manifold and valve cover, then the cylinder head, . . .

  8. Nat says:

    “And others just look at the way things used to be, are deeply happy that neither they nor anyone else has to go through that any more, and turn their attention to the sorts of problems that are on the current edge of feasibility.”

    My grad school PI, who was a student of Greg Petsko’s, loved to tell us about how he spent a year (ca. 1980) sequencing the gene of the protein he crystallized, and how he solved the heavy-atom substructure by hand. He never expressed anything other than relief that his students could do the same process effortlessly in a couple of days. I always thought this attitude was the difference between “real scientists” and people who are just screwing around.

  9. MoBio says:

    Agree with much of above. As someone who crystalizes and studies GPCRs it’s clear that XFEL and LCP crystallography will continue to provide the most useful information for drug design and discovery. I’ve not yet seen any high resolution GPCR structures via cryo-EM or even AFM.

    What we need is high enough resolution to (a) unequivocally see the ligand and side-chain orientations (b) waters and (c) other ions (typically sodium). The typical cryo-EM resolutions are not sufficient for this.

    Certainly if it worked I/we would be using it all the time.

  10. Professor Electron says:

    How much is the realistic amount that one needs for crystallization trials for a small molecule? I’ve heard people claim 0.5mg, but I’m skeptical.

    1. LW says:

      0.5 mg is doable if you’re lucky and your protein crystallizes readily. Realistically you’d probably want at least 2-3 mgs to do a full range of trials.

  11. Phil says:

    This is why I call bullshit on the argument that the time to earn a PhD needed to go from 4 years to ~6-7 because the low-hanging fruit has all been picked. A crystal structure that was once someone’s PhD thesis in the days of slide rule calculations can now be solved in an afternoon. If work that took 4 years of full-time effort using technology available in 1950 or 1900 or whenever was considered an acceptable thesis, there’s no good reason why 4 years of work can’t be an acceptable thesis today.

  12. surbhi jain says:

    Truly agreed to it. By knowing the insight structure of the protein by X-ray crystallography, we will be able to find the mechanism of the protein in reactions with various spectroscopic techniques. X ray crystallography is the basis of multiple research.

Comments are closed.