Thoughts on the STAP scandal

I was excited when my Nature table of contents email came and I saw the headline Stimulus-triggered fate conversion of somatic cells into pluripotency.  In short, scientists came up with a method, called STAP, of dunking some regular cells into an acid bath to turn them into stem cells.  Given how little is known about stem cells and their potential to make enormous advances in medicine, this could have been a revolutionary discovery.

Unfortunately, it was too good to be true.  Not long after the paper was published, other scientists came out saying they were unable to reproduce the results.  People immediately pointed fingers at the lead author Haruko Obokata.  A bad study could have resulted from a mix-up in the lab, but all confidence was lost when it came out that twenty pages of Obokata’s PhD dissertation were taken from an NIH document.  After much scrutiny, two weeks ago she agreed to retract the paper.

In the latest development in the story, one of the co-authors of the paper has confirmed that the original cells and the induced stem cells cannot have come from the same mice.

The STAP scandal has undermined Obokawa’s research institute, Nature journal and the peer-review process, and generally the way science is done.  Some have painted the incident as a gender issue, suggesting that she falsified the results in order to get ahead in a career where there are few women.  In any case, the bigger concern here is plagiarism and making up results.

I am not writing this to point fingers and chastise Obokata.  I think the incident is actually an example of science gone right.  There have been at least 10 independent attempts to reproduce the results of the study since it came out.  It highlights the need for open access publishing, so more people can read, critique, and reproduce studies (NB Nature only made this paper publicly available after much hounding).  Asking questions and being skeptical is the whole premise of science.  Imagine how many results published in smaller journals with restricted access go unchallenged.

Journals need to take this even further by requiring some level of audibility.  In this case, Obokata’s lab notes were incomplete, leaving no trail for her to defend her work.  Had she been required to submit a paper trail, falsifying results would have been difficult or impossible.  Computational journals are moving towards reproducibility by changing their policy to require authors to make code and data publicly available.  There will be fewer false findings if journals hold authors accountable and let information flow freely.