Two flow cytometry panels in a 2018 Frontiers in Immunology paper by authors from Sweden and China appeared to share some data points. The image duplications were very suggestive of post-experiment image alteration. Yet the editors accepted the authors’ excuse that it was an “accidental error”, and published a correction. For this, they will be awarded the fourth “This Image Is Fine” Award.
The problematic image
The paper with the image-of-concern is by LinLin Niu et al., published in Frontiers in Immunology (2018), DOI: 10.3389/fimmu.2018.01940 [original PDF].
The authors are affiliated with several laboratories in Chongqing, China: the Huazhong University of Science and Technology, Wuhan, China; and the Karolinska Institutet in Stockholm, Sweden.
The image in question was a flow cytometry image, Figure 3J. It appeared to have some unexpected repetitive features, pointed out to me by a reader with excellent image forensics skills.
Since the top two panels in Figure 3J showed flow cytometry plots from two different mice (WT and KO), they would be expected to show unique measurements. Indeed this is the case for most of the data points in these two panels. However some of the dots — circled here by me using yellow and orange boxes — appear to be exactly the same in both panels.
Unexpected similarities
Such partially identical panels are very unexpected. Flow cytometry plots contain lots of individual dots that should be unique, even if one were to run the same sample twice through the same machine.
If a researcher mistakenly inserted the same graph twice into their paper, one would expect the panels to look exactly the same. Not, as here, just partly the same.
Or if a sample was run through the machine and the counter was not reset afterwards, the next sample would have shown the points from the first plot, with additional points added from the second sample. Not, as here, resulting in two plots that share some data points, but miss some of the others.
Unless a flow cytometry expert can offer a better explanation, these plot similarities suggest that at least one of the images was altered after being generated in the lab.
A correction to fix an accidental error
Despite the very unexpected image similarities, the editors decided not to retract the paper, but to give the authors a chance to send in a new version of Figure 3.
In September 2020, the paper got a Correction, which reads:
“In the original article, there was a mistake in Figure 3J and Figure 4R as published. We used the wrong flow plot of KO thymus in the upper panel of Figure 3J and Figure 4R due to an accidental error. The corrected Figure 3 and Figure 4 appear below.“
Unfortunately, the correction does not include an explanation of how the two flow cytometry panels of Figure 3J ended up containing some of the same dots. The authors just said “Oops, wrong plot, here is a new one“, and the editors were totally fine with that.
The original PDFs and Figure files were silently replaced by new versions.
And so, for allowing authors to correct a potentially photoshopped image with a better version, the editors of Frontiers in Immunology have earned the fourth This Image Is Fine Award.
Updated award logo by Jon Cousins.
Science Integrity Digest 
“Such partially identical panels are very unexpected.”
Well, no. I’m not a flow cytometry expert, but one gets to these plots through a sequence of gatings (see: https://www.bio-rad-antibodies.com/flow-cytometry-gating-strategies.html). It would be very easy to accidentally wind up with a plot that is partially identical if one of the earlier gates (i.e. on dimensions not displayed on this pair of axes) was set differently for the same underlying set of data. This would let through most of the same cells, but might remove some, or add others, giving you exactly what you see here.
And if you’re working with so many samples, it may be quite easy to click on the wrong one when you’re processing each in turn. To say “The image duplications were very suggestive of post-experiment image alteration” is exceptionally uncharitable.
LikeLike