Skip to content

Genes-R-Us — Latest news on extreme longevity — or is it?

Genes-R-Us — Latest news on extreme longevity — or is it?

Posted: Wednesday, January 25, 2012 5:00 pm

Sometimes what does not get printed is more important than what does gets printed. Exaggeration sells more than detailed scientific caveats. And scientist can be so long-winded when all you really want is a short, simple story. 
Take, as an example, the popular news stories published last week sporting glittery titles like, “New blood tests predict chances of living to 100.”
Many science writers today simply do not have the background to understand the genomics research on which they report.  They are in fine company, too. Many researchers are hampered by their own lack of expertise in the evolving mathematics to keep their results from getting yanked in this briskly moving area of genomic research. 
Here is the background. The original study entitled, “Genetic Signatures of Exceptional Longevity in Humans” was published in the crème de la crème journal, Science, in July 2010. On July 2011, the paper was formally retracted due to a number of unintentional errors, including some mathematical shortcoming pointed out by experts in mathematical genomics. 
Last week, on Jan. 18, 2012, the corrected and revised version of the article, with additional highly credentialed authors from Yale, was published, spawning the aforementioned stories with the glittery titles.  
The idea of the study was to analyze DNA from 801 centenarians, with ages ranging from 95 to 119 years old, and from another 914 healthy normal individuals with average age of 75, using some common subset of SNPs to differentiate between the two groups. SNPs are those single point differences in DNA between us.
Scientists developed their best predictive formula to classify an individual as a centenarian or not. Then they would test it on two new but smaller sets of data with both centenarians and “normals” to check how well those same SNPs could differentiate the between the two groups for new populations of individuals.   
U.S. census data shows a 1 in 5,000 chance for any of us to live to be 100 years old. To attain an age of 110 years is one in 7 million. 
Let’s examine the claims in the popular press.
Although it is certainly possible to calculate SNPs from blood samples and it would be possible to build a diagnostic chip with the final 281 predictive SNPs used in the revised paper, no company in their right mind would do so.
The problem is that the prediction was just not good enough to be of any use — seriously. 
A good prediction would only classify centenarians as centenarians and classify those that are not centenarians as “normals.” Math gurus call these two concepts sensitivity and specificity, respectively. A perfect prediction aims to achieve 100 percent sensitivity and 100 percent specificity. 
The final prediction formula got 85 percent of the centenarians correct but only 26 percent of the “normals” or non-centenarians correct, using the data from the original population of 801 centenarians and 914 “normals.”
For the two smaller testing groups, some stats were somewhat worse and others were a bit better — but an increase in getting more “normals” into the correct category cost getting more centenarians wrong. 
Now let’s suppose that I went to have this (currently non-existent) “blood test” done. I am very likely to hear that I am going to live to be 100, even though the odds are very small that I really will. It’s like building a test for some medical condition that tells people they have it when they don’t.
Now there were several important results to come out of this study — but they were mostly missed by every column I read. 
More than 80 percent of the genetic variation is NOT explained by the prediction formula. This is a HUGE result. That means that a very large number of genetic variants (SNPs and other DNA features) are necessary to predict extreme survival. 
No one SNP was the key to extreme survival. The most significant SNP, rs2075650, overall could be tossed and accuracy was dipped by only 1 percent. 
Out of a thousand centenarians, only one had the risky version of this particular SNP. The risky version of this SNP, A, is strongly associated with late-onset Alzheimer’s disease and cardio-vascular disease. We can say that if someone is destined to be a centenarian, he or she will likely NOT carry the risky version of this SNP. 
Inheriting the risky version of the SNP from both parents is very rare, occurring in only 3 percent of the population.  Hence, it has poor predictive value for the general population. 
At this point the knowledge of the optimal genetics for extreme longevity still belongs to God. But for some of us, even a morsel of insight into this process is awe-inspiring.
Editor’s note: Nancy Miller Latimer has worked in scientific research and development for 27 years. Published in The Messenger 1.25.12

Leave a Comment