Friday, August 10, 2007

The Kundu case: What if ...


Update: In this long post, it is easy to lose oneself in the multiple links (and links within links). So, let me cite here "the featured link" that really clinched the issue for me: Rahul's GIF animations of seven -- yes, seven! -- sets of figures. If you want to dig more, you can follow the links on his page, as well as those at the end of this post.

Here's a quick summary (follow the links at the end of the post for details): Two papers published in 2004 and 2005 in the Journal of Biological Chemistry (JBC) by Gopal Kundu and members of his group at the National Centre for Cell Science (NCCS) have been under a cloud because of allegations that several experimental figures have been used more than once with different labels. These allegations were investigated during August 2006 - April 2007 by three different external committees (in addition to an internal one).

The very first external committee to investigate the case returned a "Not Guilty" verdict. This committee, headed by G. Padmanaban, former Director of IISc, also enjoyed an "official" status, as it was constituted by NCCS itself. Instead of settling the issue once and for all, its clean chit to Kundu has to now contend with the diametrically opposite verdict from the other two committees. After Current Science allowed its space to be used by all the key players in this controversy to air their side of the story, the debate has entered the public domain.

Last month, Rahul used very clever and expressive GIF animations of the offending figures. His work allows us to see the evidence for ourselves in all its live action glory, and the conclusion is inescapable: the figures within each set are extremely, astonishingly, and oh-so-improbably similar. In particular, I urge you to check out the sixth example to see how dramatic the similarity is -- across two different sub-figures!

Rahul has also presented his analysis in a letter to Current Science:

A thorough forensic analysis would take into account a more realistic estimate of the horizontal shifting, the vertical alignment, the positions of streaks, smears, spots, and so on – all of which are evident to the naked eye, and all of which line up as soon as one aligns the black borders. Even if every one of these metrics has one chance in three of lining up, the probability of all of them doing so would quickly become infinitesimal, even for a single image.

However, as already noted, the presence of nine duplicated images makes such an effort unnecessary. If we unthinkingly estimate that a single image has a 5% – or even 10% or 20% – chance of being genuine, the probability of all nine being authentic is already vanishingly small.

For the record, Padmanaban has responded to Rahul's letter. And Rahul has a follow-up note on his website as well.

* * *

I think it is safe to say that this game is up for Kundu. Unless he introduces some dramatically conclusive evidence.

* * *

This allows us to step back and look at a couple of "what if" questions. The first is this: what if Kundu was alerted by, say, a reviewer of the 2005 paper about the similarities between some of its figures and those in the 2004 paper. Would he have challenged the reviewer's alert as strongly as he has contested the allegations? By his own admission, the papers' scientific merit is unaffected by the suspect figures, since they form only a small part of the work. Is it possible, is it just possible, that he might have chosen to change the problematic figures with those from other experiments? Could he have chosen a similar path when the original allegations arose? Could he have opted to "correct" the paper through an erratum, instead of denying that the offending figures were the same?

The second, and the more important one, is this: what if NCCS had taken the journal JBC into confidence, and asked the latter to appoint a representative to be a member of the Padmanaban committee? After all, these are not just some results doing the rounds within NCCS; they had already been sent out into the big bad world! And it was only a matter of time before JBC decided to take another look at the paper. So, it makes sense to get all the players on board the investigative committee so that its conclusions are final and binding on everyone. This way, the matter would have seen a clean closure.

Now, look at the consequences for NCCS of the current closure-less status in which Kundu is "officially" not guilty, but his JBC paper stands withdrawn. NCCS has the obligation to do everything in its command to fight his case with JBC -- in court, if it is needed -- to get the paper reinstated. Not doing so would imply that the Lab is not sufficiently interested in protecting the interests of its scientists.

On the other hand, if Kundu loses his case in the hearts and minds of biologists (which is very likely, given the evidence that is now in the public domain), and if the perception that he is being 'protected' takes root, it can only do long term damage to NCCS. Honest scientists would think many times before wanting to work there.

Clearly, NCCS is staring at some bleak choices.

* * *

Links:

The Society for Scientific Values maintains a web page on its own investigation of the Kundu case. This page has links to the 2004 and 2005 papers from Kundu's group, and to newspaper stories on this case following the release of SSV's report.

Current Science opened up this affair to the public in its issue dated 10 June 2007. Several letters have appeared in a subsequent issue. Rahul's letter, and Padmanaban's response appear in the latest issue.

2 Comments:

  1. pradeepkumar said...

    If Kundus group had used Molecular Dynamics (Now part of GE) phosphorimager, an instrument very common and which I daily use, to scan their Western blot gels, its not that difficult to find out wheather or not they have provided the same images for two different experiments. The simple thing to do use is to open the raw file in Image Quant (the program supplied with the machine) and find out the date and time on which gels were scanned. If you double click on the auto radiogram a GEL INFO window will pop up and that exactly will give you the information you are looking for. If both the gels shows the the same time and day of scan and if both the images have same counts (some thing like pixels which the program calculates) then there is no doubt that both the auto radiograms are absolutely same. But if they had purposefully scanned the gels after different times of exposure, it will be slightly more complicated to figure out the similarities. But still it is possible. Let Kundu or Padmanabhan or NCCS provides the original auto radiogram files (the raw data from the machine) for a third party scrutiny to validate their claims. Any discussion using TIF or JPEG figures from JBC publication will have little credibility than the same discussion after comparing the raw data from machine.

  2. Rahul said...

    That's a valuable comment. One has to ask why the Padmanaban Committee did not do this already, or did not report doing this.