Thursday, September 29, 2016

Cell Press, scientific fraud, replication, and retractions

As per Retraction Watch, Cell Press will not be retracting two papers that were flagged as problematic after one author claimed to have manipulated his data to fit the desired conclusion. That author, Dr. Yao-Yun Liang, is conveniently unavailable. The first author (Xia Lin) and the corresponding author (Dr. Xin-Hua Feng) already have a retraction under their belt for "inappropriate data manipulation" found in an earlier paper, but Baylor College of Medicine (where the work was carried out) conducted an investigation, and found no evidence supporting claims of fraud in this case. So far, so good.

Now for the weird part: Cell Press had Dr. Feng get some of his friends to attempt validate the results, which they did for the Cell paper in question, and now claims that means no fraud was committed (as does Baylor College of Medicine, which uses this result to bolster their claims from their investigation). There was an interesting discussion of this over at DrugMonkey's blog a couple of weeks ago, just after the first editorial note was issued by Cell (September 8) regarding this paper. Even weirder, apparently the validation results of the Molecular Cell paper in question were inconclusive, but Cell Press won't be doing anything anyway!

This is beyond bizarre. First of all, whether the results replicate has no bearing on whether fraud was committed. We all like to think we have good scientific intuition, and sometimes that is actually true. It doesn't mean we get to publish papers with data manipulated to support our good intuition. If there was fraud, the paper should be retracted, even if the conclusions end up being sound.

Second of all, if Cell Press is going to use the "if the data replicates, it isn't fraudulent" argument, they should at least be consistent! From my understanding this is what happened:

     1) Dr. Liang says he manipulated his data in these two papers.
     2) Dr. Feng denies the allegation, and says Dr. Liang is trying to to hurt his career with these lies.
     3) Baylor College of Medicine investigates, and finds that it is a "he said-he said" problem, and says there is no evidence of fraud.
     4) Cell Press decides that Dr. Feng should get some people to replicate the result to "prove" they were not fraudulent (WHAT!).
     5) It all works out in the end for Cell, so they say that since the results replicate, it doesn't matter.
     6) It all doesn't work out in the end for Molecular Cell, but they say it doesn't matter anyway.

Huh? Something is so off with this scenario. Setting aside the fact that manipulated data does not have to be inconsistent with actual experiments (it just has to be falsified), if replication is supposed to settle the issue, then why is Cell Press ignoring the "inconclusive results" for the data in question in the Molecular Cell paper?

To my mind, Cell Press had four options:
  • They could say they will use the results of the Baylor College of Medicine investigation, and not retract. 
  • The could say the Baylor College of Medicine investigation was not thorough enough, and do their own investigation, then make a decision.
  • They could believe Dr. Liang and retract.
  • They could not know who to believe, and retain the "Expression of Concern", keeping it attached to the papers, and putting in all the information about the confession, the denial, and the Baylor College of Medicine investigation.
Any of those things would have been reasonable (if possibly controversial) responses to the allegations of fraud. Instead we got false logic about how "if it replicates, it must have been real" which was ignored when that became inconvenient for Cell Press. And people wonder why Retraction Watch is so busy?

No comments: