The media loves to tout scientific breakthroughs, and few are as toutable – and in turn, have been as touted – as CRISPR. This method of targeted DNA excision was discovered in bacteria, which use it as an adaptive immune system to combat reinfection with a previously encountered virus.
It is cool on so many levels: not only is the basic function fascinating, reminding us that we still have more to discover about even simple organisms that we thought we knew so well, but the ability it grants us to remove and replace any DNA of interest has almost limitless applications in both the lab and the clinic. As if that didn’t make it sexy enough, add in a bicoastal, male-female, very public and relatively ugly patent battle, and the CRISPR story is irresistible.
And then last summer, a bombshell dropped. The prestigious journal Nature Methods published a paper in which the authors claimed that CRISPR could cause many unintended mutations, rendering it unfit for clinical use. Havoc duly ensued; stocks in CRISPR-based companies plummeted. Thankfully, the authors of the offending paper were responsible, good scientists; they reassessed, then recanted. Their attention- and headline- grabbing results were wrong, and they admitted as much, leading Nature Methods to formally retract the paper this spring.
How did this happen? Shouldn’t the editors at a Nature journal know better than to have published this in the first place?
Alas, high-profile scientific journals publish misleading and downright false results fairly regularly. Some errors are unavoidable – that’s how the scientific method works. Hypotheses and conclusions will invariably be overturned as new data becomes available and new technologies are developed that allow for deeper and deeper studies. That’s supposed to happen. But that’s not what we’re talking about here. Nor are we talking about obvious offenses like outright plagiarism. We’re talking about mistakes that are avoidable, and that still have serious ramifications.
Two parties are responsible for a scientific publication, and thus two parties bear the blame when things go awry: the scientists who perform and submit the work, and the journals who publish it. Unfortunately, both are incentivized for speedy and flashy publications, and not necessarily for correct publications. It is hardly a surprise, then, that we end up with papers that are speedy and flashy – and not necessarily correct.
“Scientists don’t lie and submit falsified data,” said Andy Koff, a professor of Molecular Biology at Sloan Kettering Institute, the basic research arm of Memorial Sloan Kettering Cancer Center. Richard Harris, who wrote the book on scientific misconduct running the gamut from unconscious bias and ignorance to more malicious fraudulence, largely concurs (full disclosure: I reviewed the book here). “Scientists want to do good science and want to be recognized as such,” he said. But even so, the cultures of both industry and academia promote research that is poorly designed and even more poorly analyzed. In Rigor Mortis: How Sloppy Science Creates Worthless Cures, Crushes Hope, and Wastes Millions, Harris describes how scientists must constantly publish in order to maintain their reputations and positions, to get grants and tenure and students. “They are disincentivized from doing that last extra experiment to prove their results,” he said; it could prove too risky if it could cost them a publication.
Ivan Oransky and Adam Marcus founded Retraction Watch, a blog that tracks the retraction of scientific papers, in 2010. Oransky pointed out that blinded peer review – the pride and joy of the scientific publishing enterprise – is a large part of the problem. “Pre-publication peer review is still important, but we can’t treat it like the only check on the system. Papers are being reviewed by non-experts, and reviewers are asked to review papers only tangentially related to their field. Moreover, most peer reviewers don’t look at the underlying or raw data, even when it is available. How then can they tell if the analysis is flawed or the data is accurate?” he wondered.
Koff agreed that anonymous peer review is valuable, but severely flawed. “Blinded review forces a collective view of importance,” he said. “If an article disagrees with the reviewer’s worldview, the article gets rejected or forced to adhere to that worldview – even if that means pushing the data someplace it shouldn’t necessarily go.” We have lost the scientific principle behind review, he thinks, which was to critically analyze a paper. But instead of challenging fundamental assumptions within a paper, reviewers now tend to just ask for more and more supplementary data. And don’t get him started on editors. “Editors are supposed to arbitrate between reviewers and writers and they have completely abdicated this responsibility, at every journal. They do not judge, and that’s a real failing.”
Harris laments the wasted time, effort, and resources that result when erroneous ideas take hold in a field, not to mention lives lost when drug discovery is predicated on basic science findings that end up being wrong. “When no one takes the time, care, and money to reproduce things, science isn’t stopping – but it is slowing down,” he noted. Mistaken publications also erode the public’s opinion of legitimate science, which is problematic since that opinion isn’t especially high to begin with.
Scientists and publishers don’t only cause the problem, though – they may also provide the solution. Both camps are increasingly recognizing and dealing with the crisis. The self-proclaimed “data thugs” Nick Brown and James Heathers use pretty basic arithmetic to reveal statistical errors in papers. The microbiologist Elisabeth Bik scans the scientific literature for problematic images “in her free time.” The psychologist Brian Nosek founded the Center for Open Science, a non-profit organization dedicated to promoting openness, integrity, and reproducibility in scientific research. The Nature family of journals – yes, the one responsible for the latest CRISPR fiasco – has its authors complete a checklist to combat irreproducibility, à la Atul Gawande. And Nature Communications, among other journals, uses transparent peer review, in which authors can opt to have the reviews of their manuscript published anonymously alongside the completed paper. This practice “shows people how the paper evolved,” said Koff “and keeps the reviewer and editor accountable. Did the reviewer identify the major problems with the paper? Because there are always major problems with a paper.”