A paper that appeared to find evidence for a new mechanism of genetic regulation has been challenged by an analysis released today.
In May, a team led by Vivian Cheung of the University of Pennsylvania in Philadelphia reported that in an analysis of 27 people, it had found 10,210 sites where transcribed RNA differed from an individual’s corresponding DNA sequence. The finding was startling because it implied that there might be an as-yet undiscovered mechanism of ‘RNA editing’ that could disrupt the central dogma, the process whereby DNA is faithfully transcribed into matching sequences of RNA, which are translated into proteins.
But other researchers pointed out that Cheung’s team had not performed some crucial analyses to ensure that it was actually observing mismatches rather than genetic sequencing errors or accurately transcribed regions of DNA. Today, Daniel R. Schrider, Jean-Francois Gout and Matthew W. Hahn of Indiana University in Bloomington report evidence that undermines Cheung’s claim. Hahn’s team reports that 55% of the apparent mismatches reported by Cheung’s team are in fact accurate transcripts of the underlying genetic code, and thus do not represent RNA–DNA mismatches.
These transcripts appeared to be mismatches because they are highly similar to, but slightly different from, other genes. Such similar but slightly different versions of genes, called “paralogs,” are widespread in the human genome, and can differ among individuals. They are created when genes are duplicated and mutate over generations.
Cheung’s team missed these paralogous regions of DNA because the team compared its transcribed RNA sequences to the DNA sequences of one particular set of protein-coding genes, called GENCODE. But GENCODE does not contain the DNA sequence of every gene paralog.
Cheung’s team did not compare its RNA transcript data to the most recent version of the human reference genome, or to DNA sequences from the 27 individuals included in its study. Hahn’s team made these comparisons and found that just over half of Cheung’s purported RNA mismatches actually match a paralogous gene sequence in one of the 27 individuals included in the original Science paper. Thus, Cheung’s team was fooled into thinking it was seeing RNA-DNA mismatches that did not exist, Hahn says.
“Our results suggest that by failing to consider the totality of genomic data from their sample population and paralogous loci present in the human reference genome, Li et al. grossly overestimated the amount of RNA editing in the human genome,” Hahn’s team writes today in PLoS ONE.
Hahn adds that although his team is only reporting definitive evidence that 55% of the apparent mismatches are incorrect, more of them are probably incorrect, but this was difficult to prove because there is not complete DNA sequence data available on the 27 individuals who were examined in Cheung’s original study. If more complete data were available, Hahn’s team estimates that 90% of Cheung’s reported mismatches would be found to be incorrect.
And when the 55% of apparently mistaken mismatches are removed from Cheung’s data set, the remaining mismatches appear to conform to patterns of RNA editing that had been previously reported, undermining support for the idea that an unknown mechanism of RNA editing exists, Hahn says.
“I think there could be some unknown mechanism of RNA editing, but it’s certainly not as widespread as the original paper made it out to be,” Hahn says. “But the onus of showing that the central dogma is wrong is really on the original publication.”
Hahn’s paper also cites a recent paper in Nature Genetics that reported the existence of 1.809 RNA-DNA differences in 15 Korean individuals. “They did the right study of the right kind and got results that are much more believable,” Hahn says.
Hahn says his team did contact Cheung’s group, but did not receive a reply. But this will almost certainly not be the last word on Cheung’s original paper, as other analyses of the work are underway. We have requested comment from Cheung and will update this post if we receive it.
Update: Cheung’s response to Hahn’s paper, and his reply, is here.