The U.S. Office of Research Integrity is investigating work done at a Caltech lab after researchers there couldn’t replicate it, and retracted a paper based on the findings.
Here’s the notice, which ran this week in the Proceedings of the National Academy of Sciences (PNAS):
Retraction for “Selective cell death mediated by small conditional RNAs,” by Suvir Venkataraman, Robert M. Dirks, Christine T. Ueda, and Niles A. Pierce, which appeared in issue 39, September 28, 2010, of Proc Natl Acad Sci USA (107:16777–16782; first published September 7, 2010; 10.1073/pnas.1006377107).
The undersigned authors wish to note the following: “Anomalous experimental results observed by multiple members of the Pierce lab during follow-on studies raised concerns of possible research misconduct. An investigation committee of faculty at the California Institute of Technology indicated in its final report on this matter that the preponderance of the evidence and the reasons detailed in the report established that the first author falsified and misrepresented data published in this paper. An investigation at the United States Office of Research Integrity is ongoing. The undersigned authors hereby retract this paper and sincerely apologize for the inconvenience caused to other investigators.”
Robert M. Dirks
Christine T. Ueda
Niles A. Pierce
Venkataraman, who did not sign the notice, is no longer at Caltech. He’s not named among the alumni of the Pierce lab, but is in group photos there. He is listed as scientific and executive manager of the Harley Street Fertility Clinic in London.
Caltech tells us:
Later lab results did not replicate the published data.
We asked if the investigative report would be released, and Caltech responded by saying that they were cooperating with ORI’s investigation.
According to the paper:
This work was funded by the National Institutes of Health (NIH) National Cancer Institute (R01 CA140759), the Pardee Foundation, the National Science Foundation (NSF) Molecular Programming Project (CCF 0832824), the Caltech Center for Biological Circuit Design, the Beckman Institute at Caltech, the Caltech Innovation Initiative, and a Caltech grubstake fund.
The authors disclose that they hold a patent on the method used in the study. It’s not clear which of the first several on this page is the one connected to the work.
The paper has been cited 34 times, according to Thomson Scientific’s Web of Knowledge.
VERY annoying notice. They removed the PDF of the paper, and the HTML version (http://www.pnas.org/content/107/39/16777.full) has most of the figures redacted from it.
Doesn’t matter, because Google scholar brings up the PDF (get it while you still can).
http://xa.yimg.com/kq/groups/19147026/1531986034/name/RNA+selective+cancer+therapy.pdf
There’s nothing I can see in the figures that would indicate misconduct, but maybe my eyes are not good today.
Quote (from above): “The U.S. Office of Research Integrity is investigating work done at a Caltech lab after researchers there couldn’t replicate it, and retracted a paper based on the findings.”
Nothing to see here – unless you actually break out the pipettes and try replicating the experiments…
The odd thing about the gel images in Figure 2 A is that each gel image appears to be composed in 3 parts. In Adobe Reader, using the “Select Tool” (generally the deafult tool when you open a doc), click on the ladder in any one of the three gels (HCR1 HCR2 or HCR3). The Select Tool will highlight the image in blue, and you can then copy it etc. Generally, clicking on a figure image selects the entire image. Not so here.
On each of the three gel images of Figure 2 A, there is a right side portion (three rightmost lanes 7, 8 and 9); an upper left portion (upper portion of lanes 1 through 6) and a lower left portion (lower portion of lanes 1 through 6). The darkness of the bands, and the crispness of the dark edges of the bands in the upper left portion of each gel (lane 3 upper) is decidedly darker and crisper than the surrounding or lower portion of the lanes. The upper portion of the bands in each ladder also appear very dark and crisp.
Figure 5 A gel image appears to be comprised of four image components: upper and lower portions of lanes 1 through 8, and upper and lower portions of lanes 9 through 13. Again, the darkness of the bands and the crispness of the lane edges in the upper image portions looks odd indeed, compared to the fuzzy ghostly nature of the rest of each lane.
The Select Tool highlights and selects the entire gel image when clicked on the gel images in Figure 6 C.
All of the data bar graphs look odd indeed, with bars either very short or standing all the way up to 100%. Error bars poking up over 100% are puzzling to this statistician, and reflect poor understanding of proper statistical techniques applied to these data, if indeed there was any real data to begin with.
As I posted on In The Pipeline…the gel does not seem suspect to me, and I think using the select tool is a poor way to analyze this. I suggest using more thorough tools in photoshop:
A better way to do it would be to look at the fine details of the image by adjusting the levels in photoshop or some other tool.
The selection tool anomalies may just be an artifact of rendering the pdf or how the image was pasted in to the final document. If you look at the high quality pdf at PNAS and extract out the image you can see that there is substantial fine detail in all of the lanes of the gel that is continuous for each sample. Little, if any, evidence of tampering.
In fact, if you look at the first gel (HCR1) you can see that Adobe allows for a pre-defined selection of row 7,8 and 9. However, in the image I posted below, there is a clear imperfection (highlighted in red) in the background of the gel that spans row 6 and 7. This would imply that the “pre-defined selection” from Adobe is not two gels put together, as their background imperfections are consistent.
http://i.imgur.com/TrBw3.jpg
Now, it is very likely that other things were altered. I do not think the gel is suspect though.
As my new obsessions are flipping between this site and science-fraud.org, this case seems to be a good example of how to handle misconduct properly.
So many of us have co-authors who provide data we cannot possibly vet. By getting out ahead of the ORI, one can at least salvage a reputation, career and distance themselves from the fraudster.
Don’t get me wrong it is still going to sting, but I cannot believe the complicity in the levels of misconduct that are now being exposed on the internet.
Dr. Auktepus applauds.
Oh my god, I haven’t worked in the Caltech public relations office since 1992, but I’m still cursed with having to repeat over and over “One word, lower-case t, no punctuation.” It’s Caltech, not CalTech or Cal Tech or Cal-Tech or Cal-tech.
Fixed, thanks Bob.
That was fast!
From his profile at http://www.lightstalkers.org/suvir-venkataraman —
“I am recently finished my PhD in Bioengineering at Caltech and am continuing research on cancer therapy as a post-doctoral scholar in the same place. Photography is one of my long standing passions and you can see some of my work on my blog.”
From the same profile: “Last login: over 3 years ago”