Retraction notice on:

Degenhardt T, Rybakova KN, Tomaszewska A, Mone MJ, Westerhoff HV, Bruggeman FJ & Carlberg C (2009) Population-Level Transcription Cycles Derive from Stochastic Timing of Single-Cell Transcription. Cell 138, 489-501.

In this paper we used computational approaches to show that the transient transcriptional cycling observed in cell populations, i.e. periodic assembly of transcription factors and their cofactors and the resulting cyclical accumulation of mRNA, may stem from stochastic timing, sequential activation of transcription, and synchronous activation of individual cells.

Results from experiments we performed supported our model. We have recently uncovered instances in several figures where experimental data points were inappropriately manipulated by one of the authors. The manipulations do not affect the computational modeling. We have now reanalyzed the experimental data and find that they no longer provide statistically significant support for our claim of population-level transcription cycling. We are therefore retracting the paper. We sincerely apologize for any inconvenience this might cause.

Further comments:

In this publication, we highlighted an explanation for the transcriptional cycling that had been reported by a number of groups in the literature. The essence of the story was that if all individual cells in a population are activated at the same time, and if transcriptional activation is an irreversible, ordered sequential, multi-factorial process, transient oscillations with time periods of around an hour are in keeping with known physical chemical and biochemical parameters. In the same publication, we showed substantial experimental data confirming the oscillations at the levels of promoter activity, chromatin immune precipitation, mRNA and DNA looping.

A number of weeks after the publication, we were alerted by a kind and attentive colleague that two bands in figure 2D of the publication are identical whereas as they should correspond to different experiments. This induced us to go through every detail, retrieving the original experimental data and reconstructing the presentations of the experiments in the publication. Because we were able to work as a team of all the authors, we have now been able to do all this.

This team has come to the following conclusions:

1. Although most of the experimental data points presented in the publication are real, some of them have been manipulated by one of the authors, such that the resulting 'data sets' together were more in keeping with the occurrence of the oscillations.

2. Experimental figures on the basis of the true experimental data do not provide evidence of statistical significance, neither for oscillatory behaviour nor for the absence of such behaviour.

3. The figures and tables resulting from the modelling aspects are all in good order.

As a team we have therefore decided to retract the experimental parts of this work. In discussion with the journal, this has led to a retraction of the publication as a whole, again with consent of all authors. All authors stand by the modelling results of the paper, however.

We obviously feel very bad about this.

Firstly, readers of the publication may have been misled by the erroneous experimental figures and tables in our publication. We sincerely apologize for this. We point out, however, that there are a significant number of other published experimental reports of this type of transcriptional cycling including out own work on VDR. We think that also in our system there were oscillations, but vis-à-vis noise, oscillations are hard to demonstrate with statistical significance.

Secondly, there are now incorrect experimental data in the literature, i.e. the experimental data on PPARd cycling on the PDK4 gene in HEK cells of our publication. The official retraction may not be enough to remove all their traces. We shall attempt to identify colleagues who make use of the experimental parts of the publication and attend them to the problem.

Thirdly, this could hurt the image of science. We try to limit the damage by sorting out what has happened to the finest details and by being fully accountable.

And then, we also feel bad that this has happened in one of our laboratories. We have reconstructed how and why this has happened and we are putting procedures in place to prevent recurrence. The grave mistakes were made by a PhD student in training. We shall be instructing future PhD students explicitly as to what is and what is not, proper data processing (the present data manipulation was in the grey zone between non-professional outlier elimination and data creation). We shall also limit the competitive pressures to which our PhD students are subject. Although we expect our post-docs to have been trained in what are and what are not proper procedures in science, we shall also repeat the instructions for them. At the same time we are discussing standard procedures one should put in place in research groups to prevent all this from happening, e.g. was in which group leaders can identify illegitimate data manipulations.

We have prepared a document, in which we have constructed the figures and tables of the original publication side by side with the corresponding figures and tables, now on the basis of the true experimental data. This document is not a publication but will be available upon request as a confidential document.

As a final note in this letter, we emphasize that in as far as we can see, our junior co-authors Rybakova, Tomaszewska, Moné and Bruggeman hold no responsibility with respect to what has happened. The person, who actually committed the improper data manipulation, was a PhD student at the time under the supervision of one of us. She was subject to intensive time pressure, for various reasons. This is an explanation but not an excuse for what she has done, however. More recently she has worked with us loyally in order to get everything to the surface. The two of us (CC and HW) have not committed any act of data mal-manipulation. However, we are co-responsible, because we were the leaders of the groups that were involved in the study as a whole. We have not been attentive enough and have been too focused on interpreting the data as presented to us, rather than checking whether those data corresponded to actually experimental data.

Again we express our sincere apologies to the scientific community and to you as a colleague.

Carsten Carlberg and Hans V. Westerhoff