Trouble down at t'mill

This post was earlier cross-posted at Leonid Schneider's site, hence the unfrivolous tone. The version there is improved by Leonid's editing, background details and frame-story.

At the risk of working the RetractionWatch side of the street, I am here to introduce readers to a stream of retractions and corrections emerging from Cellular Biochemistry and Physiology, where the new publishers and editorial team are grappling with the toxic waste-dump buried in the Augean stable back-issues which they acquired when they took over the journal from Karger. No end of ingenuity went into composing face-saving rationales for each paper's depublication that avoid the use of words like "made-up" or "faked".

I particularly like the explanation given by or on behalf of Yu et al (2018), who claim to have belatedly realised that repeated-measures ANOVA would have been a more appropriate form of analysis, providing more robust statistics.

The authors of the article ... have requested retraction on the following basis:
In this study, they explored the effects of silencing of LFA-1 on the hippocampus of rats at different time points (0h, 24h, 48h, and 72h). The correct statistical analysis method should be repeated measurement ANOVA, while they misused the one-way ANOVA due to the limitation of our understanding of ANOVA, which might lead to some inaccuracies in this research.
The editors appreciate that the authors have asked to retract their manuscript to avoid misleading other researchers.

The implication is that the rats they decapitated and sliced up 24 hours after surgical simulation of a stroke were the same rats as those decapitated and turned into microscope slices 48 hours after the injury, and again after 72 hours, bringing them back to life each time in the manner of a horror movie. I have concerns about this but the important point is the authors' altruistic self-sacrifice "to avoid misleading other researchers".

In the Corrections, various authors provide replacement images. Missing are explanations of how the original, now-derogated images came to be manipulated before their unintentional inclusion in the manuscript, but the editors evidently decided against making an issue of it.

And as recently promised, Pan et al. (2018)

have requested retraction on the following basis:
Recently, they found that some of the results of WB and apoptosis experiments in this paper may have some errors, and some experimental results need to be verified by their team again. They apologize for the possible laxity of their paper. Therefore, they apply to retract the paper to avoid adverse effects and misleading other researchers.

Here "Found that some of the results may have errors, and need to be verified by their team again" is a term of art meaning "We outsourced the experimental aspect of our manuscript to an outside contractor (and also, it may be, much of its preparation to meet the English-language, conventional-structure expectations of the journal)." The contractor abused their naive confidence and provided images recycled from other papers, linking them all into a far-flung network of scholarly fabulation.

#13 Bao-Long Pan

Dear PubPeer, Thank you for your comments. This paper is the only one we have ever done with a third party. The name of the third party in this paper is "Nanjing Huazhong Biotechnology Co., Ltd. (Nanjing, China). We do not know the researchers who did the experiments in their team. Due to the lack of cooperation experience, we really believe the testing ability and integrity of the third party only through their email introduction. After the cooperation completed in 2017, we did not contact again because there was no other cooperation demands. Thank you two experts very much for pointing out our mistakes. Thank you two experts very much for your hard work. During this period, we are extremely distressed by the occurrence of this incident. We are also trying to safeguard our rights actively. Thank you again for your letter. Best Regards！ Bao-long Pan 2020.8.18

I have no idea of the status of these retraction requests and editorial edicts. Do they have any official standing if Karger does not recognise them? They may be 'infelicities' in the terms of Austin's speech-act philosophy and I instructed Siri to order all the popcorn while we find out what becomes of them.

I also told Siri to start Powerpoint and create this montage of papers who were outsourced to the same studio, and consequently display the same impression of criss-crossing blades as if from a melee of invisible swordfighters. More examples consequently appeared but I cannot be bothered updating the montage.

The context here is the invasion / migration method for quantifying metastatic tendencies in cancer cells from some specified cell-line, subjected to various signaling-pathway treatments. Both assays entice the cells to migrate amoeba-style through a well of nutrient gel to a membrane, where they proliferate into colonies that can be stained and counted through a microscope by long-suffering grad students. The Migration assay graduates to Invasion if the cells are further cajoled into squeezing through pores in that membrane in their eagerness to spread their disordered DNA, so that the colonies are its the far side. None of this provides any meaningful prediction of how well the treatments will control malignancy and virulence of actual cancers in actual patients... but "do-able" is more important than "meaningful" and as with animal testing, cancer research is all about looking for dropped keys under streetlights.

The broader context is the activities of a contractor / papermill, whose manifestations have metastatised all through the biomedical literature. This post is on the way of a sequel to an earlier report, comprising various out-takes.

I surmise, anyway, that each of the different cell-lines used in Migration / Invasion assays forms colonies with its own distinctive morphology, depending on the tissues to which they once belonged, and recognisable to an expert. Sometimes there are invisible swordsmen, and sometimes clusters of blue-purple leopard spots.

One long list of papers use figures characterised by a vaguely crozier-shaped colony in one of the panels. There are few explanations for how these coincidences could occur, and they all diminish the likelihood that the putative experiment actually took place. Figs 9 of Yang et al (2015) and 7A of Xu et al (2016)...



...the former marked up to show the overlaps with Fig 7A of Xiong et al (2017).

So many panels overlap even within Fig 7A of Xiong &c that they can be tiled to reconstruct a larger image. Three panels of Fig 5A from Liu et al (2017) turn out to be details of that composite.

Most recently Fig 3(b) of Gan et al (2019), and 4C of Wang et al (2019), although in the latter the palette has shifted dramatically.

Figs 5F and 5H of Xie et al (2017) provide further incarnations and overlaps.

Taking a different tangent from Xie &c, here is Fig 5B, marked up to show the overlaps with Figs 6 and 11 from Li et al (2015). I find these scenes immensely restful.

Lumpy entities are suspended in a flat picture plane in serene indifference to one another's existence, like the hieratic bird figures in Bill Hammond paintings, like the Balinese shadow-puppet theatre with a cast of balloon animals. Presiding over the revels is an orca, portrayed in the Pacific North-West style... I leave it to experts to weigh in on assigning it more precisely to the Haida or Salish or Tlingit artistic traditions.



In fact 5B is a convenient entry-point to a whole cycle of overlapping images. Just saying, by the way, that if a fire-fighter was repeatedly first on the scene to extinguish house-fires of unknown origin, one would seek a record of their movements. By the same token, suspicion turns upon PubPeer contributors 'Dendrodoa Grossularia' and 'Actinopolyspora Biskrensis' who reported many of these cases. Anyway, here a helpful montage ties in 5B with Qiao et al (2017) and Ba et al (2017).

The trail stretches in both temporal directions, back to Figs 5 and 7 of Xu et al (2015) and forward to Fig 6A of Zhang et al (2019).

We should not neglect Figs 11 and 12 of He et al (2016) despite the absence of color, for there is a lot going on.


It is difficult to show all the connections among Bai et al (2016) (Fig 4A), Li et al (2015) (Figs 7 and 12), and He et al (2016) (Figs 13 and 14) without running rapidly out of colored pencils, and glossing over the 90° rotations and horizontal flips involved in some of these correspondences. Zhao et al (2016) and Guo et al (2018) obtained their own illustration by retouching details from the same panels, but bored now.

Going back to Qiao et al (2018), the final author agreed in a PubPeer comment that there appears to be a problem though it isn't his.

#8 Alfred King-yin Lam

I agree with the first and corresponding author comments as well as the comments of in this group

The first and corresponding author will investigate in his team and have contacted the journal for the article

The first author blamed the duplications within and between Figs 8 and 9 on a student's inexperience in file management, and promised a thorough investigation to support the paper's conclusions. Our confidence in this thoroughness would be greater if the author showed more concern for the previous and subsequent occurrences of the same images.

#11 Bin Qiao

#10

Thank you very much for the question you pointed out. My students was in charged of the part of the Transwell experiment and the lab responsible for this result. As the students lack of experience, the stored names of the groups after taking pictures are not standardized and the layout is negligent, which leads to the situation that the pictures are overlapped, and the confusion of the experimental data and picture management leads to the possible picture leakage. I attached great importance to this issue. We are verifying the experimental results and contacted the editor of the journal magazine to deal with this problem. We realized the that the result may be not rigorous, and I am negotiating with the editor in chief about the retraction of the manuscript. Thanks for your patience and we will handle it as soon as possible.

It would be tiresome to go into the details of a third recurring portfolio of images, characterized by the presence of a shark in one of them (or two sharks, or none, depending on the effort put into customising the images for individual clients). But dead horses won't flog themselves, so here are Figs 4B from Bai et al (2016); 13A from Lu et al (2018); 3a from& Wang et al (2017); 6B from Zhang et al (2017); 5 from Zhou et al (2016). Some of these papers should be familiar as participants in the 'crozier' and 'orca' cycles. I omitted 4A from Chen et al (2016) on account of lack of sharks.




The point here is to rehash the evidence that we're seeing a 'contractor' papermill. Laboratories by the dozen have tabled versions of these panels as evidence that experiments occurred as they describe -- though with different cell-lines, subjected to different treatments -- and these coincidences make most sense if we postulate an independent studio that researchers hire to conduct or fake the experiments that they are not in a position to conduct or fake themselves.

The "shark" panels also provide an excuse for a couple of music videos. Do not play both concurrently for I cannot be responsible for the outcome.

Puffball spores from Oglaf

So all this began with a few recurring apoptosis flow-cytometry scatterplots, reminiscent of dust-bunnies or stipple-drawn spore-bursting puffballs, and some Western Blots where what appear to be sewing machines designed by Philippe Starck had been assembled in grids to form ransom notes. I wondered whether the spreadsheet of cases would reach the low dozens in number, enough to make the inquiry interesting (for sufficiently elastic values of 'interest').

Then things got out of hand, and new signatures of the papermill's activity came to light, which is why the spreadsheet now contains 200 entries, distributed across 55 journals. These include Journal of Cellular Biochemistry and J. Cellular Physiology (both from Wiley), with 26 and 27. By combining those two specialties, Cellular Physiol. & Biochem. managed to attract 54 (three other Karger journals - Cytogenetic & Genome Research, Gynecologic & Obstetric Investigation and Kidney & Blood Pressure Research - only managed 2, 1 and 1).

The broad spectrum contrasts with previous spreadsheets of papermill productivity which typically show more activity concentrated in fewer journals. This reflects the difference in market niche: a contrast between mass-producing paper-shaped artifacts that target certain journals which are known to have abandoned all standards, then selling the manuscripts to would-be authors as a "Your Name Here" deal with the predestined home as part of the package; versus the more collaborative model where the nominal authors have more idea about what the manuscript should be about, and whom it should cite, and which journals it should target.

On the topic of papermill signatures, the presence of hand-stippled apoptosis scatterplots or hand-drawn cell-cycle histograms are suggestive of that provenance, but authors might fake them unaided. They are only dispositive when cloned copies find their way into multiple papers, as in Spiny Norman's cousins, or these images of supplicant beavers imploring the clouds for rain.

People are really crap at imitating stochastic distributions, where the points are supposed to come from an underlying pattern, but random variables perturb the location of each point, with no continuity from one to the next. We space them too regularly, avoiding the kind of pile-ups that randomness creates... or we lean in the other direction and arrange them in linear ley-line strands. Papermill scriveners have even less motivation than usual to convincingly simulate a random process, hence the firework displays and the strangely crystalline appearance of some of these Figures. Perhaps some science-integrity hero can write a utility in R that tests the nearest-neighbour statistics for a list of coordinates, extracted from a PDF. Not just flow cytometry scatterplots; it would also be useful for the made-up regression plots of miR / protein expression / disease biomarker correlations that are so prevalent in fake cancer research.

As for the genre of hand-drawn flow-cytometry histograms, it reached its pinnacle in these depictions of cathedrals in flames (Yao et al, 2018).

Unless you prefer your church steeples all bunched up at the base, as if they are doing sit-ups to show off their abs, or beset with alien amoebas climbing them.



None of this should distract us from the emergence of other alien alphabets in which Western Blot ransom notes might be composed. In one, the protein blobs are reduced to horizontal bars, airbrushed to geometrical perfection. In another they appear to be rows of croissants, or sausage-links fresh from the meat-grinder.

This is all work in progress, though. It seems a fair guess that if a paper has WBs in the 'classic' style of Starck sewing-machines and another in the sausage-link form, AND an apoptosis dust-bunny, then both styles are the work of a single studio, though perhaps more than one specialist.

My main concern here is that all these variants on the fake-WB theme present the proteins as advancing along the electrophoresis gel at the same speed in every lane, always advancing with military precision perpendicular to the bands, so that lanes stay equally wide while the blobs retain the shape of their original wells. There are no backgrounds. This is electrophoresis porn, readers, a phrase that I never expected to find myself writing. It is creating unnatural unreachable standards, and I worry that reviewers in these journals are rejecting genuine manuscripts because the illustrative WBs display imperfections and irregularities and do not meet these blot-porn expectations.

Alert readers might wonder about a possible link between this hypothesised atelier and the prolific 'tadpole' or 'zany sardine' factory. Tadpole production peaked in 2019, so although that factory was reported earlier here (and at ScienceIntegrityDigest, and through more academic channels by Byrne & Christopher), it was active more recently than the 'Contractor mill'. Notably, both are fond of a gray-tone palette option for plotting bargraphs: Stonehenges of gray, without darker defining borders. These graybow bargraphs are not a necessary feature, but are reason enough for close scrutiny of any paper displaying them. In contrast, though, the fabrication of flaw cytometry in the tadpole oeuvre is less sophisticated and even less convincing. Thus the weight of evidence favours them being rival companies, although I am open to the possibility that a single operation transitioned from the Starck-Appliance and Sausage-link fonts for assembling WBs, to the Tadpoles style, where the bands remain linear and regularly spaced but at least reveal a background (albeit the same background in every band).

Borrowed from ScienceIntegrityDigest

BONUS RECYCLED ILLUSTRATION. Figs 5C from Liu et al (2017), 5 from Wang et al (2017). These are "colony formation assays" or "clone-forming" or "clonigenic assays". No cell mobility is involved; cells are simply sprinkled across the agar like seasoning and left to proliferate. There has indeed been cloning. Then 4B from Ba et al (2017), Fig 7A,C of Wang et al (2019), and 4 from Li et al (2016).

BONUS BLUE LEOPARD SPOTS: Wu et al (2017) and Lu et al (2018).

BONUS RETRACTION from Cell.Physiol.Biochem ("To avoid misleading other researchers"): Piao et al (2018).

Finnish folklore tells us that the tech-bro culture hero Ilmarinen forged the Sampo in all its multicolored complexity to prove his fitness to take a beauteous maiden from Pohjola as his bride. The Sampo was a magical mill or quern that ground out products conducive to the prosperity of its possessor: salt, or flour, or gold coins. "Probably nanotech," you say, ignoring the folklorists who associate it with the Grottasöngr.

But Finnish legend (in the form of the Kalevala) is a mythic artefact in its own right, an instance of the Sword-Broken-&-Forged-Anew motif: it was stitched together by Elias Lönnrot from scattered scraps of oral tradition, and filtered through his presuppositions. It is my firm belief that the fragmented opus that eventually coalesced into the Kalevala was originally an epic poem about academic advancement and publication, and the Sampo was originally a papermill... a clever software package that could write scholarly manuscripts on any chosen topic.

I must also report that the editors of Bioscience Reports are issuing so many Expressions of Concern, I suspect that they reanimated Marcel Marceau and recruited him to join the team.