Skip to main content

Life in the Drug Labs

Voodoo Nominations

I’d like to open up the floor for nominations for the Blackest Art in All of Chemistry. And my candidate is a strong, strong contender: crystallization. When you go into a protein crystallography lab and see stack after stack after stack of plastic trays, each containing scores of different little wells, each with a slight variation on the conditions, you realize that you’re looking at something that we just don’t understand very well.
Admittedly, protein crystallography is the most relentlessly voodoo-infested territory in the field, but even small-molecule crystals can send chills down your spine (as with the advent of an unwanted polymorph). For more on those, see here, here, and here, and this article. Once you start having to explore different crystallization conditions, it’s off into the jungle – solvent (and a near-infinite choice of mixtures of solvents), temperature, heating and cooling rates along the way, concentration, stirring rate, size and material of the vessel – all of these can absolutely have an effect on your crystal formation, and plenty of more subtle things can kick in as well (traces of water or other impurities, for example).
To give you an idea, with a relatively simple molecule, fructose was apparently known for decades as the “uncrystallizable sugar”. Eventually, someone sat down and brute-forced their way through the problem, making concentrated solutions and seeding them with all sorts of crystals of related compounds (another black art, and how). As I recall, the one nucleated with a crystal of pentaerythritol crystallized, giving the world the first crystalline fructose ever seen. Other conditions have been worked out since then (in crystallization there are always other possible conditions). But that’s an example of the craziness. Does anyone have a weirder field or technique to beat it?

55 comments on “Voodoo Nominations”

  1. TeddyZ says:

    I think you hit the nail on the head here. I got nothing to add.

  2. Wavefunction says:

    “Does anyone have a weirder field or technique to beat it?”

  3. Pete says:

    I agree that there is a ‘dark arts’ element to crystallization but ‘voodoo’ can have another meaning in science. Following on from ‘voodoo economics’ (numbers just don’t add up) you can get ‘voodoo thermodynamics’ (enthalpy optimization, Ligand Efficiency etc). I even did a blog post on voodoo thermodynamics which is linked as the URL for this comment

  4. AQR says:

    An example of how weird crystallizations can be, I once had a crystalline sample of a compound but was unable to crystallize its enantiomer. After a series of increasingly desperate attempts, I tried seeding with crystals of the opposite enantiomer and the sample crystallized beautifully. I don’t think this was just a fluke, since I told the story to a colleague who was facing a similar situation and it worked for him also. Any thoughts why this worked?

  5. Calvin says:

    Crystallisation is a great one but metal cross coupling still have a bit of witch-craft about them. 99% of substrates work just fine with the old standard conditions but a subset just don’t play ball. I had a Suzuki reaction that just wouldn’t go (the first time it worked was because of contamination it turned out) and naturally that was the substrate that led to the active compound. Turned out I needed anhydrous conditions, stoiciometric palladium(!) and just the right silver salt. So very expensive witch-craft. I may as well have scraped something off my shoes to get it to work. I’d point out that a pretty large pharma company tried to change the conditions by running one of those massive automated reaction optimising runs (multiple runs) with no joy. At all. Complete witch-craft. Ultimate solution was to change to a slightly different substrate and hey presto it worked under bog standard conditions. And was more active to boot. So yeah, I’ll still give the good old Suzuki a voodoo medal on its day.

  6. Anonymous BMS Researcher says:

    Biology can be even more fun. I doubt crystals care whether a man or a woman is working on them. Animals sometimes do care.

  7. Sofia says:

    I think #3 brought up a strong contender:
    Enthalpy optimization can get pretty crazy considering the amount of water molecules you need to disrupt and the consequences of perturbing them around a binding site. It is like playing Jenga in the dark.

  8. Anonymous says:

    Don’t forget Zero-g crystallization.

  9. Henry's cat says:

    During my PhD I had some great successes crystallising out my compounds. These were all small molecules and I still have some crystals as big as a fingernail. These were great for wow-factor but the ones that produced workable crystals after great perseverance were the most celebrated.
    My local prestige within the department led one day to a knock on the lab door. There were two of us in a four-person lab at the time and after we shouted ‘come in’ we were confronted by an extremely attractive female student who asked ‘who the crystal guy was’.
    Turned out she was a textile student who wanted some crystals for her final year show.
    Alas it was not to turn in to something akin to the Big Bang Theory, as the conversation quickly went south after she realised that I wouldn’t be able to magic up a few hundred thumb-sized crystals of a multitude of colours. In two days.
    I just couldn’t catch a break.

  10. anonao says:

    Agree with Derek as well. It’s a field where you can get plates with already some set of conditions, relies on anecdotes from previous phd, post-doc, on PI flair, but may end end up getting crystals because of an impurity or an air-con malfunction.
    And crystals come in so many forms as well, getting crystals may not be enough to get data (nice needles, but too thin, start again!).

  11. Magrinho says:

    Enthalpy-Entropy-Heat Capacity-Solvation Energy
    The components of my ‘mirage of understanding’.

  12. Quintus says:

    Nice reading here on problem polymorphs.
    Bučar, D.-K., Lancaster, R. W., & Bernstein, J. (2015). Disappearing polymorphs revisited. Angewandte Chemie (International Ed. in English), 54(24), 6972–6993.

  13. Dave says:

    Semiconductor material fabrication can also be somewhat of a black art, due to the ratios involved of the dopants (versus contaminants), especially for binary/trinary/quaternary semiconducting compounds, and especially as they involve crystallization into HUGE single crystals.

  14. Pete says:

    I really like the Jenga in the dark analogy, Sofia (#7), and the crux of the problem is that molecular interactions in water are non-local (with respect to ligand and protein) and the contribution of an intermolecular contact to affinity (or the other thermodynamic quantities associated with the intermolecular association) is not an experimental observable. This doesn’t inhibit self-appointed ‘experts’ from proclaiming that the enthalpy comes from the polar interactions and the entropy comes the hydrophobic interactions. Does the formation of a hydrogen bond between ligand and water have entropic consequences? No worries, just carry on regardless and try not to be so negative about the seminal contributions of these brilliant thought leaders. Then there’s ligand efficiency (the metric) which leads to a physicochemical view that changes with units of concentration (Pauli might have said that it’s not even wrong) and we wring our hands about how difficult drug discovery while bemoaning spiraling attrition in the clinic. When I look at the state of thermodynamics in the drug discovery literature, I am repeatedly reminded of the atheist’s prayer (Lord, leadeth me unto the truth and delivereth me from those who have already found it). I slipped the term ‘voodoo thermodynamics’ into a perspective three years ago and have posted the link for that as the URL for this comment.

  15. Rhenium says:

    A former advisor regaled me with a story from
    “long ago” about the professor who had a tremendous lucky streak in crystallizations. Apparently in addition to having huge numbers of samples, the key was his great beard which would accidentally provide nucleation sites.
    This may be the equivalent of the old wives tale for chemistry, but who knows!

  16. CMCguy says:

    Assume talking only chemistry and not clinical realm as that frequently appears to dwell in alternate dimensions from science. From a process perspective there are many things that might seem more like voodoo to most chemists but as many are largely engineering issues they can be translated usually by that other perspective.
    Although I understand what you mean about crystallization’s with appropriate time and effort can almost always get to a crystal states in a controlled window (and just have to be leery when approach the edges that slide to other polymorphs). I would add on scale up once you get those crystals the isolation can often be an adventure with unexpected turns as filtration technology would seem to be fundamental but in practice there are many aspects involved that can impact in subtle ways to make one realize do not understand filtration as thoroughly as believed therefore would nominate filtration.

  17. Paula says:

    My days as a crystallographer remind me, how even the animal origin of seeding whisker was a variable.
    However, that said, @6 I agree biology is another strong contender. Differentiation of cells is a survival goal and a researchers bane. Single molecule studies of proteins show a wide range of structural states that react to variations in environment that lead to VERY different cellular outcomes. Sometimes, I wonder why we find enjoyment in these multivariate problems.

  18. Lousie says:

    #15, Rhenium: I always heard that story as referring to Emil Fischer. As a consequence, I like to swear on “Great Fischer’s Beard!” in the lab.

  19. Pi* says:

    3, Pete – racemic materials are normally much easier to crystallise than enaniomerically pure ones. different melting points, etc too. They fit into different space groups.

  20. Anonymous says:

    Having screened over 3,000 different buffers at 4 different temperatures to get my first protein crystals during my PhD, I concur.

  21. Crystahell says:

    In grad school, a friend was doing her masters in carbohydrate chemistry–a hellish field if ever there was one–and she had to spend a third year of her two year masters crystalizing her carbohydrate product. Her advisor would not let a single person defend their thesis without a crystallin product. Ugly work. I never envied a single person in that lab. She finally succeeded in some weird fish tank set-up her stuff dissolved in one container with a bunch of other solvent container giving off vapors, and allowing this whole thing to equilibrate over a month. That was crystallization set up 5000 or something. She had stacks of things going. Checking on things every day. Setting up a new variation. Crazy stuff.

  22. Pete says:

    QSAR modelling is another area in which voodoo crops up. It is rare for all the training data to be disclosed and sometimes the models themselves are not disclosed. Once the model has been validated it doesn’t seem to matter whether 5 parameters have been used or 50. Validation procedures can be misled when there is a lot of clustering in the training set. I have linked a post which discusses this and also touches on other themes.

  23. Ted says:

    When I started at Upjohn (in the early 90’s) my coworkers told me about a problem that had come up a few years before. They had a drug product that they had been manufacturing on scale for a few years as a single polymorph. All of the sudden, batches started to fail as a new polymorph “contaminant” appeared. I think it was something like a solubility/turbidity test that prompted the investigation – I don’t recall DSC/TGA being a very common release assay at that time.
    Eventually, every lot failed as the new polymorph was clearly the thermodynamically favored form. In desperation, the process was transferred to a European site. The new polymorph appeared within months. They moved it again to Puerto Rico – that bought them about a year. Eventually they gave up and re-registered.
    This is why I always seeded my GMP batches with a near-religious fervor…

  24. Mark Thorson says:

    You said chemistry, without restriction to organic or bio. Therefore I nominate additives for bright electroplating. Much of electrochemistry is a black art, partly because the secrets are highly commercial and closely guarded, and partly because the practitioners don’t know why they work. There isn’t a lot of published literature on bright plating. The best-known additives are weird stuff like saccharin and certain food colorings. Do you think electrochemists have any clue why they work? I asked a PhD electrochemist once, and she said these additives tend to be large, flat molecules. That’s pretty much all they know.
    Maybe you guys could come up with some super bright plating additive in your spare time. One of those could make you more money than your next dozen failed drug candidates. These are not all used for decorative platings — some critical processes in electronics and microelectronics rely on these additives for controlling the grain of electrodeposited metals.

  25. Wavefunction says:

    Agree with Pete that QSAR can often sound like anything goes. I have similar sentiments about shiny, fancy-sounding techniques like machine learning and neural networks that people are trying to apply to drug discovery. A lot of them can be (and are) used as black boxes, and entangling correlation with causation can become impossible. I think there’s a lot of potential especially in machine learning, but we are very far from reaching that potential, and the way it’s being done right now does not inspire too much confidence.

  26. anonymous says:

    @15, @18:
    “…Another story, perhaps apocryphal, attributes an important laboratory function to Fischer’s long flowing beard. It is said that when a student had difficulty crystallizing a sugar derivative (some of which are notoriously difficult to crystallize), Fischer would shake his beard over the flask containing the recalcitrant compound. The accumulated seed crystals in his beard would fall into the flask and bring about the desired crystallization.”
    -G. Marc Loudon, Organic Chemistry, 2nd ed. (1988), p. 1228.

  27. The Iron Chemist says:

    I heard a story that Ken Raymond’s group once had a great deal of difficulty crystallizing a particular compound. As is typical for these things, nothing seemed to work. One day, they found that a bug had died and fallen into one of the crystallization vials. A beautiful crystal was growing from one of its legs.

  28. Paul D. says:

    Sounds like another application for phage display.

  29. pharmacologyrules says:

    anything that is catalyzed by scratching your beard into, does certainly qualify as voodoo

  30. Mark Thorson says:

    I think some people will want to know what species of bug that was.

  31. Captain Ned says:

    Weirder field where one can still write peer-reviewed papers?

  32. steve says:

    You all have obviously ignored the true nature of crystals, which exist to channel your chakra energy. In order to achieve true crystallization you need to reverse the process and focus your prana into the nucleation site. This will imbue the growing crystal with your life energy, which you can then use to heal the sick (or at least to bilk them out of their money….). Clearly this is a much more important venture than simply using them to scatter X-rays.

  33. John Wayne says:

    I will add to your crystal power and sell an accompanying homeopathic treatment (of that mother liqueur) to compliment the chakra energy*
    *Patent pending

  34. Anonymous says:

    For my second-year seminar (~1996) I gave a talk about methods of determining the stereochemistry of polyene macrolide antibiotics. Early in the talk I mentioned that only a couple of structures had been solved by x-ray because they were hard to crystallize. At the end of the talk somebody asked me if anybody had tried using psychics to determine the structures. I said no, psychics need crystals.

  35. Pete says:

    Folk following this discussion who like their thermodynamics to be a bit ‘alternative’ might enjoy the webinar that I’ve linked as the URL for this comment. Action starts at 22:43.

  36. Rhenium says:

    @26: Thanks!

  37. WiredAl says:

    Black art? Alchemy? Here is some real crystal voodoo. How about “Crystallographic Snapshot of an Arrested Intermediate ” in Angew. Chem. 2015, 164) and the comment published last week in ACIE? Seems that the original paper actually crystallized acetate after azeotropic drying of a complex aqueous mixture with diethylether. They did not have, as suggested, some funky odd bicarbonate intermediate. Fortunately, the authors assure us in a recently published corigenda that their calculations, using the wrong structure, are still correct! How ’bout that AIMojo!

  38. xtalibur says:

    I had crystallized a family of coiled coil peptide mutants (synthesized without any problem any purified by HPLC). They were generally quite stable, but some of the mutants just didn’t crystallize.
    I really wanted the structure of one sequence, and tried resynthesizing (maybe something went wrong last time?), and repeated many times without success. Thousands of crystal trials later, one of my 20nL drop had a nice crystal in it that gave me a 1.5 A structure.
    With biological preps, we can almost always point a finger at purity but not when you have something that is HPLC pure. One crystal out of thousands of trials (the crystal probably grew since the plastic covers eventually leak, letting your crystallization drops dry to a high concentration). I probably would never have found it if it weren’t for nano liter crystallization and nanodrop machines/imagers.

  39. formerly at Abbott says:

    Pretty much the same thing happened in the late 90s when a new polymorph appeared for Abbott’s AIDS drug ritonavir. It threatened the world’s supply. There’s a great article about it in Pharmaceutical Research 18(6):859-866.

  40. formerly at Abbott says:

    …neglected to say, the solution to the problem was different- they figured out how to control the crystallization. They had to- the new polymorph was insoluble and useless, they couldn’t just re-register a new crystal form.

  41. Kaleberg says:

    steve noted: “You all have obviously ignored the true nature of crystals, which exist to channel your chakra energy”
    Judging from the stories here it sounds like crystals have channeled a fair bit of life energy from chemists over the years.

  42. eyesoars says:

    Mark @30:
    Clearly, the insect was a humbug.

  43. Heteromeles says:

    I still remember the 1990s biology grad students doing phylogenetics who had real, physical “altars to the gel god” set up in corners of their labs. When you’re working with plants, getting all the secondary compounds out of the way so that you can amplify the DNA is a real pain. For some reason, plants have chemicals that make it hard to mess with DNA. Wonder why? Then there’s the question of whether the DNA shows anything useful once you get a sequence. Generally, it doesn’t. It’s gotten a lot easier since then.

  44. a. nonymaus says:

    Here’s one: heterogeneous catalyst synthesis. Much like crystallization or electroplating, it’s a dark art practiced by proprietary alchemists to get a high-activity Pt/C catalyst. It depends on the Pt precursor, the synthesis reagents and conditions and even the carbon source in ways that are not known and difficult to study.

  45. tangent says:

    “Sounds like another application for phage display.”
    Wouldn’t it, though? I say in utter ignorance of the fields. Would that work, to randomize the capsid into say 2^30 candidates, try to grow on each one, and if one works you crack open the crystal and mass-produce that virus code. Can we cover the space needed to get the crystallizations to work, while functioning as capsid?

  46. Some idiot says:

    @27,30 (and 42; wish I thought of that one myself… 😉 )
    Another reader on these pages had a personal example of the wingtip of a moth nucleating a crystallisation… I saw it myself….!
    But really, bugs, moths, beards, whatever: the “only” thing you need is something which can help the process of self-organisation so that a nucleus can form. So it is “just” a question of finding the right scaffold. Which is why we use seeds, miniscule bits of glass (scratching glassware, anyone?) probably flakes of skin, or fragemts of hair, or, in the more exotic cases, a bug or a moth….!
    Re@4: This is not unheard of. As long as it is something that can start the recognition process moving… Obviously, with the wrong enantiomer there will be obvious differences, but there may well be some similarities. But it will certainly not work in every case…

  47. Noni Mausa says:

    I Am Not A Chemist, but get caught up in this blog on a regular basis.
    Your discussion of rogue polymorph crystals I am finding quite spooky, flashing back on Vonnegut’s Ice 9. Oh, and on prions. How on earth did the rogue form spread? An evil idea whose time had come?
    Also isn’t the dust on moth wings basically highly articulated (crystalline?) scales? Perhaps run series’ of crystallizations using samples from different species…

  48. John Wayne says:

    Poster number 42 wins this thread.

  49. Jake says:

    I have a friend that used to do protein crystallization. My wife worked in a vet’s office, and she would collect cat whiskers for them. One of the black magic tricks they would use was to gently stroke the whisker through the drop to seed the crystals.
    You don’t get much more black magic than that. Well, maybe eye of newt would be more impressive…

  50. Mark Thorson says:

    I see an opportunity for a new product. It consists of the legs and wings of several insect species, whiskers from cats and the beards of famous chemists (Derek, I have a proposition . . .), and a proprietary blend of various other stuff, all cryogenically ball-milled to a fine powder. Every crystallographer will want to have a small vial of Thorson’s Crystal Magic for those occasions when you’re really desperate. If business was good, there should be a Blend #1, Blend #2, Blend #3, . . .

  51. c'Thulium says:

    @4 and 46: only about 10% of compounds have an affinity for the same enantiomer over the other when crystallizing (so-called “conglomerates”). As a result, many of the terrible compounds we like to crystalize will be happy to buddy-up with the opposite enantiomer (“true racemates”) while shirking their own. My guess is that doping in the enantiomer readily forms a racemic crystal that then seeds the rest of the process.

  52. sepisp says:

    #44: That. I was surprised I had to scroll so far down to find heterogeneous catalysis. For newer catalysts, it appears that many results are from blind screening and trial and error. Often nobody knows exactly why a certain additive works, or why it has to be that exact element. I am happy there’s work on nanocatalysts, tethered homogeneous catalysts and atom-scale microscopy of catalysts, since that can inject some science into what is often nothing but witchcraft.

  53. U0011579 says:

    If by ‘voodoo’ we mean capricious and prone to being difficult to predict the outcome of reliably then we would all do well to avoid neglecting the elephant in the room – namely the scientific literature (particularly of the present era). Or am I being just a little too cynical/extrapolating a token too much?!

  54. toluene says:

    The blackest art has got to be virtual screening of compounds using crystal structures. For some protein classes this is a powerful technique, however, for others it is a resource sinkhole.
    Consider Histone Methyltransferases (HMTs) where there is a plethora of crystal structures available with both peptide substrates and S-Adenosylmethionine (SAM) bound.
    Now try to design/select inhibitors to bind to either the peptide or SAM binding site (or a combination of both).
    We have tried this for years with miserable failure!
    Anyone had success other than SAM analogs?

  55. Alastair says:

    #toluene i concur. Many years ago (2005) this was tried where I worked with a 0% hit rate against SET7/9. complete waste of time.

Comments are closed.