On 2014 Feb 07, Paul Brookes commented:

A correction was recently issued for this paper... http://www.plosbiology.org/annotation/listThread.action?root=78137

The correction is a good start, but is deficient on several levels. I have outlined these in detail at a blog post on my lab' website (http://www.psblab.org/?p=268), but here's the short version...

1) The splicing issue raised WRT Figure 4A has not been addressed at all. 2) The phrase "inadvertently used the wrong blots" is inadequate, since in some cases it was not re-use of blots that was problematic, it was apparent re-use of single bands within blots. How exactly does one use the wrong band in a blot? 3) The correction images provided are askew - they appear to have been scanned in from paper copies and misaligned slightly during this process. There could be any number of reasons why files were not provided in a more direct manner (i.e. straight from the software they were generated in, without inserting a print-scan step). Regardless, the provision of apparent paper scans does nothing to enhance my confidence in these data. 4) The correction contains zero explanation for why it took 7 months to achieve.

I look forward to hearing from either the authors or the journal, about these remaining unresolved issues.

On 2013 Oct 22, Paul Brookes commented:

FYI, the follow up on this...

It has been known about on PubPeer for some time, as well as tweeted to the NYT author who wrote an article all about [http://well.blogs.nytimes.com/2013/07/17/exercise-in-a-pill-the-search-continues/]("Exercise in a Pill"), and blogged about on In the Pipeline.

The journal knows this, and was emailed several times over the past 3 months, with minimal responses ("we're still investigating" etc.) My most recent emails last week (13 weeks since the original comment was removed) have been left unanswered. This raises the question as to exactly how one goes about dealing with a paper that contains problematic data? These internal journal commenting systems and blogs and other publicity don't appear to have a very rapid impact on correcting the literature, so let's hope maybe PubMed Commons will be a bit faster?

Also, I apologize for a stupid mistake made very early on in using this PubMed Commons system - in between logging in via eRA commons and pasting my original comment across from PubPeer, I screwed up and tagged the wrong paper (albeit one with an almost identical title), then I got flustered when the comment was removed, when in fact it was just PubMedCommons staff doing their job! Put it down to a combination of me being stupid and caffeine deprived. Anyway, now the comment is in its right place, thanks to an eagle-eyed editor!

On 2013 Oct 22, Paul Brookes commented:

The following comment was left on the PLoS Biology site on July 18th, and promptly removed.

Fig. 1A, LCAD panel, lanes 2-4, appear identical to Fig. 1B LCAD panel lanes 1-3. Slightly rotated and with different exposure. This is despite these samples allegedly originating from different experimental treatments. Lots of shared features including the shapes and relative orientations of the bands, and the "kink" in the right most band, the relative shapes and orientations of the bands, and the "streak" emitting from the top right corner of the 3rd band.

Fig. 1A, PGC-1a panel (top one) appears identical to the PGC1a panel in Fig. 1B. They are simply different exposures of the same image, despite allegedly originating from different experimental treatments. Lots of shared features including the white spot in the lower part of the right-most band. The same appears true for the "SUO" blot (3rd from the top) - some "noise" spots added in panel A, but there's no doubt these are the same bands.

Fig. 4A (and elsewhere), some blots are used as loading controls for other blots, but cannot possibly have originated from the same gels, because some panels are spliced (as indicated by lines), and others are not. Sometimes it is permissible to run your samples on separate gels at the same time, in the exact same order, and then do the phospho vs. total blots separately and re-unite the data at the end. Here we are asked to believe the gels were run separately and with the samples in different order, then some of the blots were spliced (presumably to remove unwanted samples) but the others were not. This is not adherent to the usual standards of data presentation for this type of experiment.

Fig. 6B. The CYTO C panel appears to be simply a darker exposure of the one above it (COX IV). Lots of shared features, most notably the bubble above the right lane.

Fig. 3A, compare the band in the right lane of the cyto C panel (2nd blot from the top), with the band in the right lane of the CYTO C blot in Figure 4C. They are both of a shape that is too similar to be coincidental.

There are numerous other problems here.... the entire paper contains 85 (!!!) panels of western blotting data, every single one of which is "letter-boxed" to show only the band of interest, and none of which are annotated with molecular weight markers. In addition, despite the common origin of most of the samples, there is variability in the properties of the bands. For example in Fig. 4A, the phospho-ACC blot has 2 bands but the total ACC blot only has one band. Why? In Fig. 5B the ATPase blot has 3 bands, but only a single band in Fig. 3B. In several other cases, enhancing the contrast of the western blots reveals that adjacent bands have completely different color histograms - some are grayscale while others are color. As such, it is difficult to believe that these bands originated from the same scanned blot images (which is a prerequisite for being able to splice together blots).

Note... these are NOT allegations of any type of misconduct. I'm just pointing out what appears to be a very sloppy attitude toward data presentation, which seems to have resulted in a number of the "wrong" western blot images ending up in the published paper. Hey, with 85 almost identical looking images to keep track of, there were bound to be a few that slipped through the net! I'm sure these can easily be attributed to mistakes during electronic figure preparation, and corrected in a manner that in no way affects the conclusions of the paper.

On 2013 Oct 22, Paul Brookes commented:

The following comment was left on the PLoS Biology site on July 18th, and promptly removed.

Fig. 1A, LCAD panel, lanes 2-4, appear identical to Fig. 1B LCAD panel lanes 1-3. Slightly rotated and with different exposure. This is despite these samples allegedly originating from different experimental treatments. Lots of shared features including the shapes and relative orientations of the bands, and the "kink" in the right most band, the relative shapes and orientations of the bands, and the "streak" emitting from the top right corner of the 3rd band.

Fig. 1A, PGC-1a panel (top one) appears identical to the PGC1a panel in Fig. 1B. They are simply different exposures of the same image, despite allegedly originating from different experimental treatments. Lots of shared features including the white spot in the lower part of the right-most band. The same appears true for the "SUO" blot (3rd from the top) - some "noise" spots added in panel A, but there's no doubt these are the same bands.

Fig. 4A (and elsewhere), some blots are used as loading controls for other blots, but cannot possibly have originated from the same gels, because some panels are spliced (as indicated by lines), and others are not. Sometimes it is permissible to run your samples on separate gels at the same time, in the exact same order, and then do the phospho vs. total blots separately and re-unite the data at the end. Here we are asked to believe the gels were run separately and with the samples in different order, then some of the blots were spliced (presumably to remove unwanted samples) but the others were not. This is not adherent to the usual standards of data presentation for this type of experiment.

Fig. 6B. The CYTO C panel appears to be simply a darker exposure of the one above it (COX IV). Lots of shared features, most notably the bubble above the right lane.

Fig. 3A, compare the band in the right lane of the cyto C panel (2nd blot from the top), with the band in the right lane of the CYTO C blot in Figure 4C. They are both of a shape that is too similar to be coincidental.

There are numerous other problems here.... the entire paper contains 85 (!!!) panels of western blotting data, every single one of which is "letter-boxed" to show only the band of interest, and none of which are annotated with molecular weight markers. In addition, despite the common origin of most of the samples, there is variability in the properties of the bands. For example in Fig. 4A, the phospho-ACC blot has 2 bands but the total ACC blot only has one band. Why? In Fig. 5B the ATPase blot has 3 bands, but only a single band in Fig. 3B. In several other cases, enhancing the contrast of the western blots reveals that adjacent bands have completely different color histograms - some are grayscale while others are color. As such, it is difficult to believe that these bands originated from the same scanned blot images (which is a prerequisite for being able to splice together blots).

Note... these are NOT allegations of any type of misconduct. I'm just pointing out what appears to be a very sloppy attitude toward data presentation, which seems to have resulted in a number of the "wrong" western blot images ending up in the published paper. Hey, with 85 almost identical looking images to keep track of, there were bound to be a few that slipped through the net! I'm sure these can easily be attributed to mistakes during electronic figure preparation, and corrected in a manner that in no way affects the conclusions of the paper.

On 2014 Feb 07, Paul Brookes commented:

A correction was recently issued for this paper... http://www.plosbiology.org/annotation/listThread.action?root=78137

The correction is a good start, but is deficient on several levels. I have outlined these in detail at a blog post on my lab' website (http://www.psblab.org/?p=268), but here's the short version...

1) The splicing issue raised WRT Figure 4A has not been addressed at all. 2) The phrase "inadvertently used the wrong blots" is inadequate, since in some cases it was not re-use of blots that was problematic, it was apparent re-use of single bands within blots. How exactly does one use the wrong band in a blot? 3) The correction images provided are askew - they appear to have been scanned in from paper copies and misaligned slightly during this process. There could be any number of reasons why files were not provided in a more direct manner (i.e. straight from the software they were generated in, without inserting a print-scan step). Regardless, the provision of apparent paper scans does nothing to enhance my confidence in these data. 4) The correction contains zero explanation for why it took 7 months to achieve.

I look forward to hearing from either the authors or the journal, about these remaining unresolved issues.