An error in the research database UK Biobank resulted in a request for the retraction of a paper from a UC Berkeley professor with implications on past CRISPR research Friday.
The paper — based on the research of Rasmus Nielsen, a professor in the UC Berkeley integrative biology department, and postdoctoral fellow Xinzhu Wei — asserted that people with the CCR5-∆32 mutation have a 21 percent increase in mortality. Originally published in June, the paper was inspired by the work of Chinese biophysics researcher He Jiankui, who claimed to have introduced a mutation similar to CCR5-∆32 in twin babies by using CRISPR-Cas9 gene-editing technology.
The mutation supposedly increases resistance to HIV infection, but Nielsen and Wei’s research suggested that it had adverse side effects as well. The specific genetic code that Nielsen and Wei analyzed, however, was missing in some of the UK Biobank participants, according to an email from Sean Harrison, a professor at the University of Bristol.
“This happens with genetic data – looking at the genetic code is tricky – but it looks like who was missing this value were not random, so bias could have been brought in here,” Harrison said in an email. “This means the survival analysis (the analysis that gave rise to the 21% increased mortality figure), as well as the Hardy-Weinburg analysis (the analysis the authors claimed supported the deleterious effect of the deletion), were biased, and so erroneous.”
Harrison, who was initially skeptical of the paper, conducted the same experiment with different methods and a different genetic code. He did not find an association between CCR5-∆32 and a higher mortality rate.
Nielsen and Harrison communicated via Twitter and email, and Nielsen and Wei repeated the experiment with the same setup as Harrison. Their results, however, still showed that the mutation resulted in a higher mortality rate.
While Harrison’s findings did not necessarily refute the paper, he did inspire David Reich, a Harvard University professor of genetics, to replicate the study, which provided enough evidence to nullify statistical support of the original paper.
The paper, originally published in Nature Medicine, was peer reviewed, but the error was not caught prior to the article’s publication.
“There was an error in the data from a public database that we worked on,” Nielsen said in an email. “We should have caught the error, but that is not something that you can expect reviewers to do.”
On Friday, Nielsen announced on his Twitter that the claims made in the paper no longer have scientific backing, as most of the research was likely based on erroneous data.
“Overall, I see this as an example of post-publication peer-review that went really well,” Harrison said in the email. “The end result isn’t brilliant – retraction never is – but without twitter and UK Biobank, the conversation probably wouldn’t have happened, and the paper may never have been corrected.”