https://www.twitter.com/ras_nielsen/status/11...3202871298
“David Reich (inspired by the work of Sean Harrison) has found an error in the UK Biobank data that likely explains most or all of our results regarding CCR5 delta-32. We will work with the Nature Medicine editors to get the publication record corrected. 2/3”
https://www.twitter.com/GaetanBurgio/status/1...1455269888
It appears the results that CCR5∆32 homozygous mutation reduces life span from the UK biobank cohort don't hold following post publication peer review and replications studies. From @RebeccaDRobbins
https://statnews.com/2019/09/27/major-error-u...ens-lives/
Major error undermines study suggesting change introduced in the CRISPR babies experiment shortens lives
By REBECCA ROBBINS @rebeccadrobbins SEPTEMBER 27, 2019
A scientific study published this past spring came with damning implications for Chinese scientist He Jiankui, who created the world’s first gene-edited babies: People with the rare genetic variants that He tried to engineer into embryos, the study asserted, had an increased death rate.
On Friday, the paper’s senior author said his study was wrong.
“The one thing that all scientists fear the most is to find out that a major result they have published was based on erroneous data,” Rasmus Nielsen, of the University of California, Berkeley, wrote on Twitter. He said that he had been notified of an error in the data from the massive genetic database that Nielsen and his collaborator, Xinzhu Wei, had analyzed to reach their conclusion.
The error “likely explains most or all of our results” regarding the genetic variant that He tried to alter, he said.
Nielsen told STAT that the error stemmed from the specific single nucleotide polymorphism, or genetic marker, that he and Wei looked at. In the U.K. Biobank data, the marker they chose to work with had systematic errors related to “genotype calling” at that site in the DNA; that’s the process by which the genotype is determined for each individual in the sample at each site.
The way the genotypes were being called caused certain genotypes to show up less frequently than they should have, Nielsen said, apparently generating the erroneous signal around increased mortality.
“So if there is an effect on mortality,” Nielsen told STAT, “it is certainly not as strong as we previously reported.” He added: “We are very sorry for the confusion we might have spread by publishing these results.”
Nielsen said that he and Wei tried to replicate their results in several other databases, both freely available to researchers and proprietary, and that work reinforces their conclusion that their published finding was erroneous. In doing that work, they also turned up independent errors related to genotype calling at that same SNP in another genetic database, known as gnomAD, Nielsen said.
Nielsen said the evidence suggests that the problem is likely specific to the marker they analyzed and “not a general sign of poor data quality in the U.K. Biobank.”
When the paper was published in June, it generated some pushback online from scientists who took issue with technical aspects of Nielsen and Wei’s approach.
One of those critics was Sean Harrison, a University of Bristol epidemiologist and statistician who has worked with data from the U.K. Biobank. Shortly after the paper’s publication, Harrison wrote in a blog post that he tried to replicate the analysis using U.K. Biobank data. When he did so, “I found… Nothing,” he wrote.
Nielsen told STAT that he and Wei later worked with Harrison to determine that the reason for the discrepancy in their analyses was that Harrison used a different SNP than they did. That led to the alarming conclusion that while the SNP they looked at generated the signal around increased mortality, that signal wasn’t present in nearby markers.
Nielsen said on Twitter that it was Harrison who inspired Harvard scientist David Reich, who has previously studied the CCR5 variant, to probe the results of the study.
Nielsen said that Reich and his team have “shown quite clearly and convincingly that the error is in the marker” that Nielsen and Wei chose to work on.
Reached by email, Reich told STAT: “I think it’s best just to have a scientific discussion and not to comment to the press while this discussion is going on.”
“David Reich (inspired by the work of Sean Harrison) has found an error in the UK Biobank data that likely explains most or all of our results regarding CCR5 delta-32. We will work with the Nature Medicine editors to get the publication record corrected. 2/3”
https://www.twitter.com/GaetanBurgio/status/1...1455269888
It appears the results that CCR5∆32 homozygous mutation reduces life span from the UK biobank cohort don't hold following post publication peer review and replications studies. From @RebeccaDRobbins
https://statnews.com/2019/09/27/major-error-u...ens-lives/
Major error undermines study suggesting change introduced in the CRISPR babies experiment shortens lives
By REBECCA ROBBINS @rebeccadrobbins SEPTEMBER 27, 2019
A scientific study published this past spring came with damning implications for Chinese scientist He Jiankui, who created the world’s first gene-edited babies: People with the rare genetic variants that He tried to engineer into embryos, the study asserted, had an increased death rate.
On Friday, the paper’s senior author said his study was wrong.
“The one thing that all scientists fear the most is to find out that a major result they have published was based on erroneous data,” Rasmus Nielsen, of the University of California, Berkeley, wrote on Twitter. He said that he had been notified of an error in the data from the massive genetic database that Nielsen and his collaborator, Xinzhu Wei, had analyzed to reach their conclusion.
The error “likely explains most or all of our results” regarding the genetic variant that He tried to alter, he said.
Nielsen told STAT that the error stemmed from the specific single nucleotide polymorphism, or genetic marker, that he and Wei looked at. In the U.K. Biobank data, the marker they chose to work with had systematic errors related to “genotype calling” at that site in the DNA; that’s the process by which the genotype is determined for each individual in the sample at each site.
The way the genotypes were being called caused certain genotypes to show up less frequently than they should have, Nielsen said, apparently generating the erroneous signal around increased mortality.
“So if there is an effect on mortality,” Nielsen told STAT, “it is certainly not as strong as we previously reported.” He added: “We are very sorry for the confusion we might have spread by publishing these results.”
Nielsen said that he and Wei tried to replicate their results in several other databases, both freely available to researchers and proprietary, and that work reinforces their conclusion that their published finding was erroneous. In doing that work, they also turned up independent errors related to genotype calling at that same SNP in another genetic database, known as gnomAD, Nielsen said.
Nielsen said the evidence suggests that the problem is likely specific to the marker they analyzed and “not a general sign of poor data quality in the U.K. Biobank.”
When the paper was published in June, it generated some pushback online from scientists who took issue with technical aspects of Nielsen and Wei’s approach.
One of those critics was Sean Harrison, a University of Bristol epidemiologist and statistician who has worked with data from the U.K. Biobank. Shortly after the paper’s publication, Harrison wrote in a blog post that he tried to replicate the analysis using U.K. Biobank data. When he did so, “I found… Nothing,” he wrote.
Nielsen told STAT that he and Wei later worked with Harrison to determine that the reason for the discrepancy in their analyses was that Harrison used a different SNP than they did. That led to the alarming conclusion that while the SNP they looked at generated the signal around increased mortality, that signal wasn’t present in nearby markers.
Nielsen said on Twitter that it was Harrison who inspired Harvard scientist David Reich, who has previously studied the CCR5 variant, to probe the results of the study.
Nielsen said that Reich and his team have “shown quite clearly and convincingly that the error is in the marker” that Nielsen and Wei chose to work on.
Reached by email, Reich told STAT: “I think it’s best just to have a scientific discussion and not to comment to the press while this discussion is going on.”
2
