Gel (see footnote at end for a brief description of gels aimed at non-biologists) splicing is a term that describes the cutting and pasting of images of lanes (where 1 lane = 1 sample) and placing the images of the lanes in a different order or even combining lanes from different gels. A more extreme form is to simply shift the subsection of the lane, corresponding to the probed molecule, from one lane to the next.
This is wrong and it always has been. However, in post publication peer review on PubPeer, it is often defended, particularly for “older” papers, from a decade or more ago. This then raises arguments about what was acceptable then and are we shifting the goalposts of scientific integrity? The matter has even been a “Topic” on PubPeer.
While I agree that space constraints on gels and in publications led to editors requesting the removal of blank lanes (an important point raised by “Unreg” in the above Topic discussion I do not think that this explains the large number of splicing events coming to light. Today I came across a very nice argument by “Peer1” on PubPeer, to which I can find no riposte – the fact that journal editors aided and abetted the practice cannot condone it. This is Peer1’s argument
This has always been the case, be it in the time of the scalpel and the camera or Photoshop.
The same applies to manipulation of any other data, such as spectroscopy or FACS. Think about it this way. You discover a fossil, with some pieces apparently not fitting the jigsaw of your hypothesis. Rather than go back to the field to search for more data, you simply get out the tools and alter the fossil.
Yes, it is hard, slow and painstaking work and always will be. Those who put their fossils together using glue and claim that this is what was found in the field rightly deserve criticism and that their paper be withdrawn, regardless of whether the actual predicted fossil turns out to exist. The one described in the paper never existed, so the paper cannot either.
I do realise that fossils are put back together again, but I think the point being made is that the joins are explicit, whereas hiding the fact that there is a join is wrong.
Footnote
Gels of agarose and polyacrylamide are used for the analysis of nucleic acids and proteins. On a gel there will be several lanes, each loaded with a different sample. This allows the comparison of biological samples. By using probes to reveal the identify of particular nucleic acids or proteins, gels allow changes in individual molecules to be monitored in complex samples (cells, tissues, even entire model organisms). So immensely powerful, and these techniques have led to important advances in our understanding of biochemistry and cell biology.
Ferniglab Blog 
As a non-gel scientist (I am more a FACS person) sometimes I do wonder if these concerns about gel splicing are a little too much nit-picking. If for instance you are looking at whether Protein X is up or down regulated by treatments X,Y,Z in one cell line. Say you did three different experiments for each of the treatments on three different gels and they show X up regulates, Y doesn’t change, and Z down regulates, what would be the problem of ‘splicing’ those three results in the one figure? As long as you have your loading control displayed for each of the treatments wouldn’t that be OK?
Having said that, from reading PubPeer I can see times where there is justifiable concerns about this splicing. But I wonder if the finding of ‘splicing’ alone is necessarily all that bad? As for problems with FACS data, well that can be left for another day…
Paul,
Happy to host a posting on FACS or link to one.
Paul, there would indeed not be a problem, as long as the authors were open about the fact that the lanes were from different gels. One reason for this is that there is variation from run to run and sample to sample, hence a preference for putting all the relevant samples on one gel, or, simply putting the three treatments as 3 panels, A, B, C!
Splicing is deception, because it leads the reader to believe all the samples were run on one gel, with a corresponding decrease in experimental uncertainty.
An allied issue is the habit (from the days of printing on paper and saving space) of presenting a horizontal slice of the gel containing the “bands of interest”. There is a very good reason why the entire gel needs to be presented: proteins are complicated. From splicing of mRNA through to posttranslational modifications, we end up with a few to hundreds of isoforms of a protein. These may be regulated differently. In addition, many antibody detection systems on Westerns will highlight “aspecific” bands, but whether these are truly aspecific or not is rarely addressed. A Western blot detects “immunoreactivity”, we then have to establish what that immunoreactivity might be. Having spent £10,000s on antibodies over the years, I know that a lot of antibodies work like partial protein stains – you get >10 bands on a gel, authors choose the one they like, then they start to splice the horizontal splices. Next step, which occurs far too often is to simply combine such horizontal splices into a work of fiction.
So simple guidelines
1. If lanes in a figure come form different gels, show that this is the case explicitly.
2. Show the entire gel, if an electronic publication, in the main body of the paper, if a hybrid paper/electronic, in SI.
Thanks for the explanation. I agree it would be best to show the entire gel, and now that we have electronic publication it should be possible. I must admit my experience with blotting is limited to doing a Northern Blot long ago for my Masters project, so I usually assume everything is fine once it passed the reviewers. But PubPeer and Retraction Watch has certainly opened my eyes up about some ‘irregularities’ that have been going on.
From PubPeer at least there seems to be less FACS data issues overall, but when there is it does look like straight forward data manipulation (duplication of results etc).
True there are less FACS data being questioned, but whether that reflects better practice or simply that it is easier to concoct we will only know when we have true open data and meta data!
[…] relating to multiple instances of gel splicing (why I think this particular practice is wrong is here). Nonetheless, the frustration of peers is kept pretty much under control, to their […]