Skip to main content

Chemical News

A Synthetic Retraction

There’s an odd retraction in the synthetic chemistry literature. A synthesis of the lundurine alkaloid core from the Martin group at Texas was published last year, and its centerpiece was a double-ring-closing olefin metathesis reaction. (Coincidentally, that reaction was one of the “Black Swan” examples in the paper I blogged about yesterday – the initial reports of it from the 1960s weren’t appreciated by the synthetic organic community for many years).
Now the notice says that the paper is being retracted because that RCM reaction is “not reproducible”. (The cynical among you will already be wondering when that became a criterion for retraction in the literature – if it works once, it’s in, right?)
There are more details at The Heterocyclist, a blog by the well-known synthetic chemist Will Pearson that I’ve been remiss in not highlighting before now. While you’re there, fans of the sorts of chemicals I write about in “Things I Won’t Work With” might enjoy this post on the high explosive RDX, and the Michigan chemist (Werner Bachmann) who figured out how to synthesize it on scale during World War II.

18 comments on “A Synthetic Retraction”

  1. Chemjobber says:

    Ivan Oransky to the white telephone, please.

  2. MR. Fixit says:

    I have a number of thoughts here. The cynic in me asks the question “did the authors embellish?”. the little person sitting on the other shoulder says “give them the benefit of the doubt.” I can speak from personal experience, Ruthenium based metathesis catalysts, from any source, can have batch to batch variability in purity, activity and side reactions. The bottom line is I can place the blame for this on any number of sources.

  3. anonymous says:

    Either they made the compound or they didn’t.
    If they did, then they should not retract. They should do a full blown investigation of source, lots, impurities, etc., until they know the source of the irreproducibility. And then publish it.
    If they didn’t, than “not reproducible” is a euphemism for “someone screwed up.” Analytical data was (willfully or not) misinterpreted in an overoptimistic way. It happens. In this case retraction is of course the correct action, but “not reproducible” is a fig leaf for something much less interesting: human error.
    If they don’t know whether they made it or not, than that’s just sloppy and shame on them.

  4. dearieme says:

    Enlighten me: when you make something that’s new and publishable, or that was made by a new and publishable route, do you store a sample indefinitely in case questions arise?

  5. anchor says:

    #3-The compounds, I believe were made as evidenced by its NMR spectra that have been posted in the supplements section of the original paper. The yields during the RCM metathesis was poor. I wonder if they did this reaction on an amide rather than amine during this low yielding critical reaction. I always reckoned that during multistage synthesis, the critical step that you are utilizing and if it is the selling point for that paper, always be high yielding.

  6. anonymous says:

    #5 Anchor:
    I agree that a high yielding key step is necessary for publication in a premier journal, but I don’t believe it affects overall publishability. That’s what TL is for. Or even perhaps OL in this case.
    If you really believe that they successfully made the compounds (and I have no reason to think otherwise based on the data, although the C13’s are noisy), then isn’t a retraction overkill?
    I guess we ought to applaud Martin for having such high standards regarding reproducibility. If the rest of the community acted this way we’d have many fewer publications. But I, for one, believe that in an empiricial science like chemistry, a little irreproducibility is OK, particularly at the boundaries of knowledge. When I attempt to apply literature methods, I don’t expect them to work the first time. I expect that I’ll need to optimize and re-jigger for my particular circumstance. That’s OK.

  7. DCRogers says:

    #3 “In this case retraction is of course the correct action, but “not reproducible” is a fig leaf for something much less interesting: human error.”
    Au contraire, explicit details of the human error are quite interesting; indeed, they are the point if we want to judge how a process failed us.
    It is a pet peeve of mine that these details are jealously guarded, making the retraction akin to the public “apologies” that never quite get around to saying exactly what they’re apologizing for.
    On a similar note, any time I hear “computer error” I sense a fallible human using a squirt of binary blame the same way a squid uses ink.

  8. Anonymous says:

    Has anybody ever wondered what type of calibrated scales these synthetic groups use that can distinguish and can measure minimum weights of 1.0 milligram. I understand they are taring flask and weighing by difference, but still measurements that result in weights less than the minimum weight that the scale can measure should not be accepted. I have a hard time accepting a result of 1.4 mgs, when most scales used in academia have miniumum weight specification at 20 mgs.

  9. Colonel Boris says:

    @ number 8: I’ve often workd with balances accurate (not just precise) to 0.01 mg, which are vital to preparing magnetism samples. Even our bog-standard Mettler-Toledo Random Number Generator balance can still manage +/- 0.2 mg if no-one’s walking past it at the time and you’re patient.

  10. barry says:

    We may never know how many different syntheses of quinine were completed in Gilbert Stork’s labs over the years. One of his post-docs estimated it at 25, of which only one (or two?) satisfied professor Stork’s criteria for publication. Other labs have other standards, of course.

  11. Anonymous says:

    @number 9: I dont doubt that your mettler toledo random number generator balance can read out =/- 0.2 mg, but what I think you are talking about is the resolution of the balance. What I am talking about is the minimum sample weight a balance can handle reliably. This means making 10 replicate weighings of a certain mass and calculating the Standard deviation (SD) associated with those weighings. If the (SD x 3)/measured value is less than 0.001, then the minimum weight the scale can handle is 3 x SD x1000. For the popular models of Mettler Toledo analytical balances used in academia ( based on models I have seen throughout my undergraduate, graduate and postdoctoral career) the minimum weight is 20 mgs. Any measurement under 20 mgs, can be read from the balance readout, but that measurement is not reliable.
    I think many of these synthetic groups use balances outside of their acceptable precision.

  12. newnickname says:

    @8 and others: I think I know what you mean. I mostly use analytical balances that read to 0.0001 g (0.1 mg) but I usually ignore the last place. I can tare a weighing paper on the balance, make a fingerprint on it and then weigh the fingerprint. There are many sources of error that can confound getting an accurate mass and yield after many manipulations and transfers.
    I sometimes weigh my products by NMR. If the standard proton (8 or 16 pulses) looks “only this good”, it’s sub-mg; if it looks “this good”, it’s around 1-2 mg. Sometimes that’s all I need to (or can) know.

  13. RB Woodweird says:

    newnickname said: “I sometimes weigh my products by NMR.”
    Here’s a thought. Partially derivatize the inside of an NMR tube with a suitable silyl group such that is has one methyl group or other group giving a singlet. Put a solution in the tube containing a known amount of protons. Run NMR, compare ratios. Now you have an NMR tube with a known amount of protons as an internal standard which you do not have to mix with your unknown.

  14. Anon says:

    @8 & @11.
    Educate yourself about analytical balances. Yes some top-loaders lack accuracy, but current analytical balances (such as the Mettler Toledo XP205) measure to 0.01 mg and the minimum weight (1.4mg) is not an issue with a difference measurement (tared flask vs tared flask plus sample).

  15. Red Fiona says:

    Re: 8 – There are a few down to 1 mg balances available. They’re somewhat jealously guarded. I say this because our group are presently working to try to engineer the return of ours that went walkabout during someone else’s lab move.

  16. Anonymous says:

    @14. I understand what you are getting at but most QA investigators may not agree with your assesment. My work consist of analyzing Drug substance for release. As part of our standard operating procedure, all balances have to be calibrated and their minimum weights determined. Any measurement that results in a weight lower than the minimum weight, reguardless if that weight is a result of a difference measurement or not, will be flagged by QA and leads to various investigations and various delays.

  17. InfMP says:

    I have an XP205. it’s sooo good.

  18. Curt F. says:

    @16. Anonymous:
    Do “most QA investigators” have a reason for disagreeing? Are your SOPs really written in a way that refused to recognize the ability of most common analytical balances to make difference measurements? If so, why? Are you trying to argue against weighing by difference (seems like a bad argument to me) or just warning folks here that the QA world has been a bit slow to catch up to the current reality and thus sometimes may question perfectly accurate procedures because they are relying on outdated SOPs (seems like a sensible thing to warn others about).

Comments are closed.