About a year ago, Acta Crystallographica Section E issued a bombshell editorial. The journal was pulling 70 papers from two groups of researchers at the same Chinese university after discovering that the structures they reported had been fakes.
As the editorial explained, the fraud was detected during a routine review of the structures by Ton Spek, of Utrecht University in The Netherlands. According to the editorial:
Initially, unexplained Hirshfeld rigid-bond alerts and unusual metal-ligand donor-atom distances led to the discovery that metal atoms had been transposed and that more than one structure had been ‘determined’ using identical sets of data. Investigation of these cases sparked a search of papers written by the correspondence authors involved [more on this in a minute].
The falsified structures have many features in common: in each case, a bona fide set of intensity data, usually on a compound whose structure had been correctly determined and reported in the literature, was used to produce a number of papers, with the authors changing one or more atoms in the structure to produce what appeared to be a genuine structure determination of a new compound. The worst example generated no fewer than 18 supposedly original structures from a single common set of data. There is nothing to suggest that the authors of the original papers describing the real structures are in any way aware of, or complicit in, this fraud.
The two main authors, H. Zhong — with 41 pulled papers — and T. Liu — with 39 — from Jinggangshan University, were fired in the wake of the scandal, Nature reported in January 2010. According to that article, Chinese researchers submitted “half of the 200,000-odd crystal structures” published by Acta E in the five years leading up to the retractions.
Now, after an extensive investigation into other papers, the journal and/or the authors are retracting 11 more articles “as a result of problems with the data sets or incorrect atom assignments.”
Peter Strickland, managing editor of IUCr (International Union of Crystallography) Journals, which publishes Acta E (and C), among other titles, told us that
The 11 new retractions … are the result of a follow up study, based on problems we have investigated since the initial retractions. These articles are by authors other than Zhong and Liu and represent cases where the reported crystal structures are wrong due to incorrect atom assignment or poor crystallographic knowledge. Thus, in these latest cases the problems were most probably unintended.
Back to those “unexplained Hirshfeld rigid-bond alerts” and “unusual metal-ligand donor-atom distances.” We asked Strickland to explain in closer to lay terms what all that means:
The checking programs we use on each submitted crystal structure look at a large number factors that may indicate how well the structure has been solved. The Hirshfeld test investigates whether the components of the vibrations of two atoms constituting a bond have sufficiently equal values in the direction of the bond as required for a rigid bond. The other test you mention looks at distances between atoms. The values are compared with usual values known from prior practice and if the values differ from what is expected, the program will note that there is a potential problem.
Lists of the Zhong and Liu articles retracted last year can be found here and here.
One of the major problems with Protein Crystallography and biophysics in general, is that very few physicists are actually involved in it. Physics bods who truly understand the methods are now very few in number as the fields have come to be dominated by molecular biologists and biochemists. Nothing wrong with this inter-disciplinary working of course but it means there is now a greatly reduced core of people with the necessary physical, chemical and mathematical know-how to spot this kind of nonsense.
Are you suggesting they may have been simply incompetent and submitted 70 papers just because they were dorks? I’m not saying it is not possible, but is it likely?
Otherwise an interesting comment, and I fully agree; I cannot name a single physicist I know of working on protein crystallography.
This journal (as I understand it) doesn’t publish protein crystallography, but crystallography of non-biological molecules. Don’t know much about this field, but it seems like it probably has more physicists than does the field of protein crystallography.
It is unlikely (IMO) that crystallography has a higher incidence of fraud, but more that it has a much greater chance of *detecting* fraud. This is because there is a large amount of Open data against which values can be checked (e.g. are they in the right range?). There are also mainly internal relations which, in a valid structure, are self-consistent and it is the failure of these which give clues that something is wrong. Crystallography has one of the highest commitements to data quality and validation.
Sorry to join this conversation 10 years’ late… Ironically, a significant number of the fraudulent Acta E papers were refereed and processed by physicists with insufficient chemical knowledge.