Upon realizing they had experienced a case of mistaken cell-line identity, the authors of a 2014 Nature paper on lung cancer think “it prudent to retract pending more thorough investigation,” as they explain in a notice published Wednesday.
The problem seems to stem from more than just honest error, according to corresponding author Julian Downward, a scientist at the Francis Crick Institute in the UK.
In a 1,215 word statement, sent to us via the Director of Research Communications and Engagement at Cancer Research UK, which funds Downward’s research, Downward told us the backstory not presented in the journal’s retraction note:
Following a series of exhaustive London Research Institute (LRI) [now part of the Francis Crick Institute] internal and external investigations, it was concluded that the data presented in Extended Figure 4 of the paper had been inappropriately manipulated, and that Dr Kumar was responsible. While no other problems have been identified in the study, it was considered that the most appropriate course of action would be for the paper to be retracted.
The paper, “HMGA2 functions as a competing endogenous RNA to promote lung cancer progression,” suggested that a particular gene plays an unusual role in the growth of lung cancer. Here’s a line from the abstract:
Taken together, these results suggest that Hmga2 promotes lung carcinogenesis both as a protein-coding gene and as a non-coding RNA; such dual-function regulation of gene-expression networks reflects a novel means by which oncogenes promote disease progression.
But the cell lines used for RNA sequencing do not line up with those presented in the legend of a figure concerning the sequencing. We noted earlier this week that cell line substitution is a pretty typical cause of problems. But this seems to be a more serious strain of the common mistake.
It was discovered, Downward said, after an outside expert reached out to him:
Shortly after the paper was published, I was contacted by an expert in the US who pointed out apparent irregularities in the cell lines and RNASeq experiments presented in the paper.
After addressing the expert’s concerns with first author Kumar, Downward says he “was not reassured that they resulted simply from experimental mishap.”
Investigations followed, according to Downward:
I brought the issue to the attention of the Director of the London Research Institute, Dr Richard Treisman, who appointed an internal Assessment Committee. The Report of the Assessment Committee concluded that there was a real possibility that Dr Kumar had committed scientific misconduct in assembling Extended Data Figure 4 of Kumar et al. The Director decided to proceed with a formal investigation of the matter, and to suspend Madhu Kumar from work forthwith.
The External Investigation Committee was appointed and carried out a detailed analysis of the case, concluding that Dr Kumar had manipulated the data presented in Extended Data Figure 4 of Kumar et al and had thereby committed scientific misconduct.
Under the LRI Scientific Misconduct procedure, Dr Kumar appealed against this finding to Cancer Research UK. Dr Nic Jones, the CRUK Chief Scientist, appointed the Appeal Panel. In addition to considering Dr Kumar’s stated case to appeal the findings of the investigation, the Appeal Panel uncovered new evidence that unequivocally supported the Investigation Committee’s finding that Dr Kumar had committed scientific misconduct.
Before the committee could write up a report, Kumar resigned, Downward said. He apparently took a job in Canada, but then resigned from that as well; we were unable to track him down. The spokesperson for Cancer Research UK said that, “To my knowledge, Dr Kumar is no longer working in research. ”
Downward said he tried to validate the data, as well:
Repetition of a number of other experiments in the paper using coded samples did not reveal any other obvious problems with the published data. These experiments included those presented in figures 1 and 3 of the paper. The experiment in Extended Data Figure 4 was subsequently repeated entirely independently by others in the lab and our bioinformatics core facility using validated cell lines and coded samples such that the identity of the samples was unknown to the experimenters. This gave data that was qualitatively somewhat similar to the published figure. Let7 was the most strongly selected for miRNA in this repeat experiment, but the strength of the Sylamer scores was about ten fold less than in the original figure, and the overall impact of the corrected figure much less compelling. However, given the misconduct findings, the inability to repeat the experiment in Extended Data Figure 4 to my satisfaction, and the potential uncertainties arising from the cell line confusion, I feel that the only appropriate course of action is to retract the paper from Nature.
Downward lays blame on Kumar alone for the mistake:
I would like to emphasise that the investigations undertaken were enormously thorough and no blame whatsoever attaches to any of the other authors of the Nature paper.
He’s also doing a back-check on some of the previous experiments that Kumar did in his lab, though no problems have been found so far:
Dr Kumar is a coauthor on five other publications from my laboratory, and we investigated whether there are problems with these papers. The investigations undertaken in connection with the misconduct allegations, included detailed analysis of Dr Kumar’s laptop and archived data, and did not reveal evidence of problems relating to other publications.
In this Letter, we reported that Hmga2 promotes lung cancer progression in mouse and human cells by operating as a competing endogenous RNA for the let-7 microRNA family. It has been brought to our attention that the cell lines used in the RNA sequencing (RNA-seq) experiment presented in Extended Data Fig. 4 of the Letter cannot be those specified in the figure legend. Moreover, these data cannot generate the Sylamer plots presented in this figure. The cell line substitution also casts doubt on which cells were used for experiments presented elsewhere in the paper. Although replication of other experiments in the Letter have not uncovered any further inconsistencies, given this uncertainty and the clear issues with Extended Data Fig. 4, we think it prudent to retract the paper pending more thorough investigation. We apologize for any adverse consequences that this may have caused.
A request for further comment sent to the Nature editor responsible for cancer papers, Barbara Marte, and the Nature editorial director, Ritu Dhand, was forwarded to a Senior Press Officer at Nature, Rebecca Walton. She declined to provide an additional comment:
Nature does not comment on retractions beyond the published retraction text.
Richard Treisman, the Director of the London Research Institute, provided this statement on the investigation that lead to the retraction:
Obviously when concerns are raised about a paper it is first of all important first to investigate the nature of those concerns, and to assess whether they might have resulted from misconduct, and if so, by whom. And then to follow that with a rigorous formal investigation of the evidence. Nevertheless, both Julian Downward and I were shocked as the results of the investigations emerged together with the findings of misconduct against Dr Kumar. Such behaviour is completely unacceptable.
All scientists dread situations of this sort. Resolving them through an investigative process is both stressful and a significant drain on resources and time which otherwise could be spent on research. I am grateful not only to those scientists at LRI who devoted time to the investigations, but also to the senior scientists from other organisations who devoted substantial time and effort to ensuring that the proceedings of the Assessment Panel, Investigation Committee and Appeal Panel were rigorous and fair.
Julian has acted entirely correctly and fairly throughout this long process, and has very delicately managed a difficult situation for both his laboratory and his research programme. He has my complete and continuing support.
The paper has been cited 46 times, according to Thomson Scientific’s Web of Knowledge.
Hat tip: Kerry Grens
Like Retraction Watch? Consider supporting our growth. You can also follow us on Twitter, like us on Facebook, add us to your RSS reader, and sign up on our homepage for an email every time there’s a new post.