Will you live to 100? In 2010, scientists published a study proposing a "genetic signature" that could help answer that question. But they retracted the paper last year when critics discovered errors that invalidated the results. Now the team is back with a revamped study that proposes a different signature; it's less predictive in most people but, they say, it still underscores that your chance of living a very long life has a powerful genetic component.
The flap began 18 months ago, when biostatistician Paola Sebastiani, geriatrics specialist Thomas Perls, and their colleagues at Boston University reported in Science that they had identified 150 gene variants that, taken together, could predict with 77% accuracy one's chance of becoming a centenarian. The work was immediately criticized by geneticists who said it incorrectly inflated the importance of a number of gene variants, making the signature look more powerful than it really was. Sebastiani and Perls subsequently said they had not realized that the microarray technology they used to identify the various gene variants had certain limitations. Last July, acknowledging that they had failed to take those into account, they retracted the paper.
At the time, Sebastiani and Perls promised to redo the work and publish any new results. They recruited additional researchers from Yale University who have expertise in analyzing genomewide association studies. While the research team analyzed the original study group of just over 1000 very elderly people—mostly centenarians but some in their 90s—and about 1200 controls, they added another small cohort of 60 extremely elderly individuals, whose median age was 107.
The genetic findings from this reanalysis were less startling than the original ones: Unlike in the Science paper, in which several specific gene variants carried a significant association with a 100-year-plus life span, here almost no variants were very significant on their own. Only when the 281 gene variants were taken all together was there a signature with enough statistical power to predict old age, the team reports this week in PLoS ONE. In most of the cohorts studied, the predictive power was also less, about 60% compared with 77% in the Science work. However, the researchers did find that the older the cohort, the more likely the genetic signature was to correctly predict extremely long life—predictability reached 85% with the oldest group. "It's very consistent with the idea that the heritability of longevity increases with age," because the power of genes becomes ever more important, Perls says.
The new work makes much more sense than what was published earlier and retracted, says David Goldstein, a geneticist at Duke University in Durham, North Carolina, and one of the early critics. "There ought to be some diffuse predictive power for some combination" of gene variants, he says, because it's "pretty clear" that longevity is passed down through generations. "What they're seeing here may reflect a real heritability of longevity."
Perls, Sebastiani, and their colleagues want to replicate the findings in other groups of centenarians and also try to better understand the biological mechanisms driving their signature. "We made a mistake," Sebastiani says. "At the same time, we continue to believe this is a good paper."
*This article was corrected 24 January to reflect that the retracted Science paper was published last year; in fact, it was published in 2010.