Home » The Daily Scan

Uncertainty and Response

July 08, 2010

Critics of the recent Science paper on longevity study say that "several things stand out as unusual from this paper," as Daniel MacArthur put it at his blog, Genetic Future. Namely, the paper shows a very large effect size and level of accuracy — especially for SNPs that hadn't been previously shown to be protective against disease — and it has the technical issue of having used different versions of a SNP chip to test the cases and controls. Study authors Paola Sebastiani and Thomas Perls took part in an online chat yesterday to discuss their work. In it, Sebastiani says that they are reviewing the data to determine the effect of using different genotyping technologies, but adds that their preliminary analysis shows it is a "limited problem" but that it is "one that has to be followed." The full chat can be seen here.

MacArthur points out that 23andMe took a closer look at this paper to determine whether the model proposed could predict longevity when applied to its customer data. At the Spittoon, 23andMe says:

The authors of the new study say their model scored a 0.93 for [a c] statistic. But when we compared 134 23andMe customers with age ≥ 95 to more than 50,000 controls, we obtained a test statistic of 0.532, with a 95% confidence interval from 0.485 to 0.579. Using 27 customers with age ≥ 100, we get a value of 0.540, with a 95% confidence interval from 0.434 to 0.645.