Adventures in peer review

The issue of inappropriate image manipulation in figures published in peer-reviewed journals has received much attention in the last few years. I often wonder whether this has led to more “ugly” results – which obviously have not been manipulated – passing through the peer review process. It is also increasingly becoming a requirement to quantify data present in western blots and agarose gels (e.g. RT-PCR results) which are, by nature, qualitative assays. Often, it seems that reviewers accept the results of such quantitation, even if the quality of the blots/gels are suspect. I come across examples of these problems regularly – in good and bad journals. The example below comes from a very recent issue of a highly ranked specialty journal.

poor western blot with misleading statistics

The authors exposed cells to Treatment 1 or Treatment 2 for different lengths of time and analyzed expression of Protein A by western blot. The Protein A blot is not terrible, even though it is not clean, and you need to take the authors’ word that the protein the arrow is pointing at is indeed Protein A (especially since there are no molecular weight markers indicated). The actin blot, however, is quite awful. Actin is a control protein whose expression should not be modulated by the treatments. The problem is that lanes 3-7 of the actin blot have clearly saturated the western blot film (they are black whereas lanes 1 and 2 are grey) indicating that actin levels are much higher in those samples and that equal amounts of protein were not loaded in all lanes. Also, the samples in lanes 3 and 4 have apparently leaked, giving a single actin band with no clear boundary between the lanes.

Because of these two issues, the authors have no business trying to quantify this result. To quantify bands in a western blot, they cannot be saturated, or you end up underestimating the amount of protein present. They are doubly wrong in trying to quantify neighboring bands whose boundaries can’t be discerned!

Below the western blots, the authors present a graph whose values represent the density of the “Protein A” bands normalized to the density of actin bands. A final issue that should have been recognized by the reviewers of this article is that the statistical analysis of the quantitation is incorrect. The authors repeated this experiment only twice, but the error bars on the graph are standard deviations, which should be based on at least three independent experiments.

This problematic figure makes up only a small part of a large and complex paper. Nonetheless, it consists of several figures that present densitometric quantitations of other gels, making the reader wonder whether they are similarly unreliable!

This entry was posted in Uncategorized. Bookmark the permalink.

Leave a Reply