B.R.S.M. Yield isn't everything


Where Did It All Go Wrong?

I’m sure life as an isolation chemist is hard. First of all, you have to actually find a source of interesting molecules, and while this sometimes involves diving in spectacular locations, or trekking through unspoiled rainforest to pick rare fruit, I’m sure it more often involves literally HPLCing shit or eviscerating four tons of eels. Furthermore, when you’ve actually got the compound, that’s only half the battle, as Nature is unbelievably creative at devising unique and surprising architectures to baffle the unwary. Synthetic chemists spend large amounts of time bewildered by NMR, and we get some pretty big clues from what we actually put in the flask to start with. Starting from scratch is even harder, even with the modern array of analytical equipment. Even the gold standard technique of X-ray crystallography isn’t perfect, and there have been some very famous natural products misassigned even with the aid of this breathtakingly powerful tool, including competition molecule diazonamide A, and kinamycin C.[1] Having said all that, looking at this recent example of a proposed natural product structure that was revised by synthesis, I have to say that I think I could have done a better job myself. Drunk.

Seriously? For starters, the structure on the left has a chemical formula of C22H18O6, which the authors somehow acquired incorrectly from HRMS data, whereas the one on the right has the formula C13H8Br4O3, with a highly uncommon splitting pattern thanks to its four bromine atoms, and is a known compound, having already been isolated from similar organisms twenty one years ago. After Kozlowski and Podlesny synthesised the original structure and noticed a few discrepancies with the reported data they did what synthetic chemists usually do in such circumstances – contact the isolation chemists. Fortunately, both the spectra used to assign the structure and a sample of the natural product itself were both available, which is better than one can usually hope for in such circumstances. Then, in a surprising turn of events they found that

‘after an analysis of the original spectroscopic data, it also appears that there are 13 peaks in the 13C NMR spectrum rather than the 11 published. These two extra peaks are very closely associated with peaks at 117.4 and 150.4 ppm’

Huh? That’s a bit odd, isn’t it? It seems they were a bit careless with proton NMR, too:

‘On the other hand, the 1H NMR data for the reported natural product (Table 1) are missing both phenol peaks and lack any of the expected splitting patterns. Analysis of the original spectrum indicates that it was obtained at high concentration, which could account for the observed broad peaks and the missing phenol signals. ‘

I can’t believe no eyebrows were raised by the supervisor and reviewers when every single aromatic CH was reported as a singlet! Yes, ortho-couplings on naphthalenes are smaller, but still pretty commonly observable! Fortunately, solving the puzzle proved easy, as once Kozlowski and co-workers obtained the correct molecular formula (by HRMS) then a literature search turned up a tetrabrominated phenyl ether with a matching formula. The data matched, and thus it transpired that all they had was a sample of a known natural product, and the binaphthyl system that drew their attention was just a figment of someone's imagination.I feel particularly bad for the student who put all the effort into synthesising a single atropisomer of the reported structure, when it was assigned so carelessly![2]


  1. In the first case, oxygen and nitrogen were confused, and in the second a diazofluorene was misassigned as an N-cyanocarbazole. Nicolaou and Snyder reviewed a bunch of examples of Molecules That Were Never There in angewandte a few years back, and you can grab the paper for free off Snyder’s website here. In a bizarre coincidence, while researching the Yonemitsu oxidation this morning, I accidently found this presentation that gives good background on the reassignment of diazonamide a.
  2. Yes, really. The isolation guys report an optical rotation, which given the true structure, seems surprising to me. See the paper for details of the asymmetric biaryl coupling used.



Comments (18) Trackbacks (0)
  1. Also see the bosutinib story: pipeline.corante.com/archives/2012/05/14/bosutinib_dont_believe_the_label.php

  2. One of my favorite tidbits from Harran’s diazonamide a story: they got so frustrated by the problems with the original structure that they printed out a copy of the original paper and burned it.

  3. that polybrominated marine natural product looks remarkably like triclosan antiseptic – I wonder if the mechanism is the same and if this brominated stuff could be used as a triclosan replacement

  4. I help teach graduate Organic Spectroscopy, and I describe my pedagogical mission as “tragedy avoidance.” We constantly struggle with synthetic chemists who are sure they know what they’ve made, and are only taking NMR and MS data to support, not test, their identity. I’m not at all surprised that the authors of the papers here would wave off the counterevidence staring back at them, and proceed to publish or allow to be published erroneous conclusions. I’m goin to use these instances as teaching materials so future chemists know the importance of accurate molecular characterization.

  5. I tried to enter a comment last night but it didn’t show up. I tried reentering it and got an error message saying it was a duplicate comment but it never appeared! What’s going on here?

    • Sorry, it got marked as spam automatically by WordPress. I have marked it as safe and it’s up now. I get around 100 spam comments a day, so I don’t check them very thoroughly any more. Thanks for mentioning bosutinib – I should have remembered that.

  6. This example is a really shocking one. Indeed there are mis-assignments in the literature all the time. In fact, we are constantly reviewing assignments in the literature for insertion into our NMR chemical shift library and we find errors on average about 8-10% of the time.

    Regarding the classic Nicolau and Snyder paper, some of my colleagues did a nice follow-up to this in Natural Products Reports in 2010. Basically, the authors took on a lot of the challenges pointed out by Nicolau and Snyder in their paper and plugged them into our CASE system to see how it would deal:


    I recently took on the bosutinib case as well:




  7. For a starter, they must have injected something completely different on their HRMS, since the found mass doesn’t even make sense for in-source fragmentation or the like (disregarding the fact that no one could oversee a Br in the mass spec like in, ever). Plus all formulas in the paper are likely wrong, or am I missing something: why would they report “HRESIMS m/z 378.1105 (calcd for C22H18O6 378.1103)”, when what you measure in ESI is either (what I would expect for this compound) [M+H]+ 379.1176 or (unlikely) [M-H]-? Or is it some kind of weird standard in the field to calculate back to your hypothetical M and publish that?

    And what is “HRESIMS were measured using a Bruker Daltonic micro TOF with electrospray ionization” supposed to tell me? Not even positive or negative? Did they somehow invent magical neutral mode mass spectrometry?

    I mean, I’m a mass spec guy, but is this not something totally obvious?

    • Yeah, the MS info is certainly a mystery! I’m also a bit mystified by the optical rotation, as others have pointed out on the In The Pipeline thread. Regarding the MS, I wondered if they maybe managed to massively contaminate the sample with plasticisers or something and then found the MW of those. My old research group had a collaboration with some isolation chemists in Malaysia, who would occasionally send us a few mgs of plasticisers to NMR, MS etc. They really didn’t seem to grasp that you couldn’t keep transferring 0.5mg of natural product using plastic syringes and containers. I only have limited knowledge of MS, but might four bromine atoms be easier to miss than one?

      • It’s weird. I’m just an undergrad so my knowledge on MS is more limited, but if I remember correctly is very easy to find Br or Cl in MS because you have to signals very close together because of the isotopes, and a very defined proportion on the intensity side.

        I have a question… on the wrong assigned molecule you have this symmetry and the molecule has this single bond joining the two parts together. Is it possible that this has free rotation and therefore they should have seen some kind of splitting pattern??? Or it’s not rotating and that’s why they expect very simple signals??

        It looks like they sent a completely different sample the second time. I mean I don’t know how heavily you can contaminate a sample so you don’t get a splitting pattern where you should (speaking of the four bromine atoms)

Leave a comment

No trackbacks yet.